GE Healthcare

Multi Target Analysis Modulefor IN Cell Analyzer 1000 Product User Manual

Code: 28-9035-38

228-9035-38UM Contents, Rev A, 2006

Contents 1. Legal and License Information 4 1.1. Legal 4 1.2. Notice to Purchaser 4 1.3. License Agreement 5 1.3.1. Software License Agreement 5 1.3.2. Grant of License for Software 5 1.3.3. Copyright 5 1.3.4. Restrictions 6 1.3.5. Software Copies 6 1.3.6. Limited Warranty 6 1.3.7. Disclaimer to other Warranties 6 1.3.8. Damage Liability 6 1.3.9. General Provisions 7

2. Introduction to the Multi Target Analysis Module 8 2.1. Overview 8 2.2. Example assays and image data 8 2.2.1. Cell viability 8 2.2.2. Cell viability and apoptosis 10 2.2.3. Cell cycle 11

3. Workflow Overview and Tips 13 3.1. Assay 13 3.2. Acquisition 13 3.3. Analysis 13

4. Initiating the Analysis Application 14 4.1. Selecting a Mode 14 4.2. Opening an Image Stack 15 4.3. Navigating an Image Stack 16 4.4. Using the Select Protocol Window 17

5. Creating an Analysis Protocol 18 5.1. Entering a Protocol Name 18 5.2. Password-protecting a New Analysis Protocol 19 5.3. Specifying the Assay Name and Microscopy Type 20 5.4. Defining Input Images 21 5.5. Using the Multi Target Analysis: Segmentation Window 23 5.5.1. Setting Segmentation Parameters on the Basis of Measurements from an Image 23 5.5.2. Nuclei Segmentation 24 5.5.3. Cells Segmentation 27 5.5.4. Organelles Segmentation 30 5.5.5. Reference Segmentation 33 5.6. Using the Multi Target Analysis: Measures Window 34 5.7. Using the Multi Target Analysis: Filters Window 38 5.8. Using the Multi Target Analysis: Define Classifiers Window 39 5.9. Using the Multi Target Analysis: Summary Window 40 5.9.1. Summary data 40 5.9.2. Subpopulation data 41 5.10. Using the Time Lapse Stack Analysis Window 42 5.11. Using the Z-Stack Analysis Window 44 5.12. Using the Data Management Window 44

28-9035-38UM Contents, Rev A, 2006 3

6. Filters and Classification 46 6.1. Classification using a Threshold filter 46 6.2. Classification using a Linear Discriminant 2D filter 50 6.3. Classification using a Decision tree filter 55 6.4. Defining the use of the Classification filter 58 6.5. Example Classification Protocols 60 6.5.1. Cell viability 60 6.5.2. Apoptosis 63 6.5.3. Cell Cycle 68

7. Using Supervised Classifiers 77 7.1. Overview 77 7.2. Preparing for annotation and classification 77 7.3. Using the annotation tool 77 7.3.1. Using the Classes page to define cell classes 79 7.3.2. Using the Annotation page to assign cells to defined classes 80 7.3.3. Using information in the Values page 82 7.4. Building a classification protocol 83 7.4.1. Choosing a classification algorithm 84 7.4.2. Selecting features for classification 85 7.4.3. Saving the Classification Protocol 88 7.4.4. Adding the Classification to the Analysis Protocol 89

8. Analyzing Images and Assessing the Results 90 8.1. Analyzing an Image Stack 90 8.2. Using Graphical Displays to Assess Analysis Protocol Performance 91 8.3. Colour-coding the Plate Map 92 8.4. Viewing Analysis Results within the Workstation Software 94 8.5. Viewing Previous Analyses 96

9. Using the Analysis Protocol Editor 97 9.1. Analysis Protocol Editor accessed through the Image Stack and Analysis application 97 9.2. Analysis Protocol Editor accessed through Analysis Protocol Manager 100

10. Using the Batch Queue 102 10.1. Using the Batch Queue 102 10.2. Creating the Batch Queue Folder 102 10.3. Opening the Batch Queue Manager 103 10.4. Adding Image Stacks and Analysis Protocols to the Batch Queue 104 10.5. Running a Batch Analysis 111

11. Glossary 114

4

1. Legal and License Information

428-9035-38UM Chapter 1, Rev A, 2006

1.1. LegalGE Healthcare and GE Healthcare Biosciences are trademarks of GE Healthcare companies.

GE and GE Monogram are trademarks of General Electric CompanyGeneral Electric Company reserves the right , subject to any regulatory approval if required, to make changes in specifications and features shown herein, or discontinue the product described at any time without notice or obligation.

Contact your GE Representative for the most current information.© 2006 General Electric Company - All rights reserved

Microsoft Excel and Windows 2000 are trademarks of Microsoft Corporation Microsoft is either a trademark of Microsoft Corporation in the United States and/or other countries

‘The IN Cell Analyzer 1000 system is the subject of patent numbers US 6563653 and 6345115 and US patent application number 10/514925, together with other granted and pending family members, in the name of GE Healthcare Niagara, Inc.

Software © GE Healthcare Biosciences Niagara Inc 2005 - All rights reserved © GE Healthcare Biosciences UK Limited 2005 - All rights reserved GE Healthcare Biosciences Niagara Inc is a wholly owned subsidiary within the GE Healthcare Biosciences group of Companies. All goods and services are sold subject to terms and conditions of sale of the company within the GE Healthcare Biosciences group, which supplies them. A copy of these terms and conditions is available on request .

The IN Cell 1000 Analyzer system is for research purposes only. It is not approved for diagnosis of disease in humans or animals.http://www.GE Healthcare.comGE Healthcare Biosciences UK LimitedGE Healthcare Place Little Chalfont Buckinghamshire HP7 9NA UKGE Healthcare Biosciences ABSE-751 84 Uppsala SwedenGE Healthcare Biosciences Corp800 Centennial Avenue PO Box 1327 Piscataway NJ 08855 USAGE Healthcare Biosciences Europe GmbHMunzinger Strasse 9 D-79111 Freiburg Germany

1.2. Notice to PurchaserThe IN Cell Analyzer 1000 system is sold for use in a variety of research applications. The purchase of this product does not include a license under any patent or intellectual property to use IN Cell Analyzer 1000 in any particular application. It is strongly recommended that the purchaser consider the need for a license to the intellectual property of others that may cover an intended use.

By using this system, the purchaser acknowledges the above referenced license constraints and accepts responsibility for all patents that may apply in using IN Cell Analyzer 1000 in any particular application.

The IN Cell Analyzer 1000 and associated analysis modules are sold under license from Cellomics Inc. under US patent numbers US 6573039, 5989835, 6671624, 6416959, 6727071, 6716588, 6620591 6759206; Canadian patent numbers CA 2328194, 2362117, 2282658; Australian patent number AU 730100; European patent number EP 1155304; Japanese patent number JP 3466568 and other pending and foreign patent applications.

528-9035-38UM Chapter 1, Rev A, 2006

1.3. License AgreementSTANDARD SOFTWARE END-USER LICENSE AGREEMENT IMPORTANT - PLEASE READ CAREFULLY

THIS IS THE LICENSE AGREEMENT THAT END-USER IS REQUIRED TO ACCEPT BEFORE INSTALLING AND USING GE HEALTHCARE BIOSCIENCES SOFTWARE. CAREFULLY READ ALL OF THE TERMS AND CONDITIONS OF THIS LICENSEAGREEMENT BEFORE EITHER (A) OPENING THE SEALED PACKAGE CONTAINING THIS SOFTWARE AND/OR (B) PROCEEDING WITH THE DOWNLOADING AND/OR INSTALLATION OF OR USING THIS SOFTWARE. OPENING THE SEALED PACKAGE CONTAINING THE SOFTWARE INDICATES END-USER’S ACCEPTANCE OF AND AGREEMENT TO BE BOUND BY ALL OF THE TERMS AND CONDITIONS OF THIS LICENSE AGREEMENT AS DOES CLICKING THE APPLICABLE “I ACCEPT” OR EQUIVALENT CONTAINED IN THE SOFTWARE. END-USER IS NOT PERMITTED TO DOWNLOAD AND/OR INSTALL AND/OR USE THIS SOFTWARE UNTIL END-USER HAS AGREED TO BE BOUND BY ALL OF THE TERMS AND CONDITIONS OF THIS LICENSEAGREEMENT. BY ACCEPTING ALL OF THE TERMS AND CONDITIONS OF THIS LICENSE AGREEMENT, YOU ALSO REPRESENT AND WARRANT THAT END-USER IS DULY AUTHORIZED TO ACCEPT THE TERMS AND CONDITIONS OF THIS AGREEMENT ON BEHALF OF END-USER AS YOUR EMPLOYER. IF END-USER DOES NOT AGREE WITHALL OF THE TERMS AND CONDITIONS OF THIS LICENSE AGREEMENT AND CHOOSES NOT TO OPEN THE SEALED PACKAGE AND/OR INSTALL THIS SOFTWARE, END-USER MAY OBTAIN A REFUND OF THE AMOUNT PAID FOR THIS LICENSE BY PROMPTLY RETURNING THIS SOFTWARE AND ITS PACKAGING IN UNMODIFIED FORMTOGETHER WITH WRITTEN CERTIFICATION THAT THE ORIGINAL SOFTWARE HAS BEEN RETURNED AND NO COPIES MADE, TO THE GE HEALTHCARE BIOSCIENCESCOMPANY THAT PROVIDED THE SOFTWARE TO END-USER NO LATER THAN 14 DAYS FROM END-USER’S RECEIPT OF THE SOFTWARE. NO REFUNDS WILL BE GIVEN FOR PRODUCTS THAT ARE OPENED OR ARE MISSING COMPONENTS CONTAINEDTHEREIN.

1.3.1. Software License AgreementThis is a legal agreement between the end-user (“End-User”) of this software product (the “Software”) and the GE Healthcare Biosciences group company (“GE Healthcare”) supplying the Software to End-User. Third party suppliers whose software has been incorporated into this Software are direct and intended beneficiaries of this Software License Agreement .

1.3.2. Grant of License for SoftwareSubject to payment of any agreed fees, GE Healthcare grants End-User a non- exclusive, non-transferable royalty-free license to use the Software on a single computer, unless otherwise agreed. If a multi-user license is agreed, End-User may use the Software on a number of computers corresponding to the number of licenses End-User has purchased. If the computer is attached to a network then End-User is responsible to make sure that the Software may only be used by a number of concurrent users that corresponds to the number of licenses End-User has purchased.

1.3.3. CopyrightThe Software is owned/licensed by GE Healthcare and is protected by copyright laws of United States and other countries and by international treaty provisions. Title to the Software (including, but not limited to originals, translations, compilations and partial copies, if any, and any intellectual property rights therein) shall not pass to End-User. End-User may not copy the written materials that accompany the Software.

6

1.3.4. RestrictionsEnd-User may not rent , lease, or sell the Software. End-User may not modify, translate, reverse engineer, decompile, disassemble or otherwise attempt: (i) to defeat , avoid, bypass, remove, deactivate or otherwise circumvent any software protection mechanisms in the Software, including without limitation any such mechanism used to restrict or control the functionality of the Software; or (ii) to derive the source code or the underlying ideas, algorithms, structure or organization from the Software (except to the extent that such activities may not be prohibited under applicable law). The Software is provided with RestrictedRights. Use, duplication or disclosure by the U.S. Government is subject to restrictions set forth in subparagraph (c)(1) of The Rights in Technical Data and Computer Software clause at DFARS 252.227-7013 or subparagraphs (c)(1), and (2)of Commercial Computer Software – Restricted Rights at 48 CFR 52.227-19, as applicable

1.3.5. Software CopiesExcept for one backup copy, you may not make copies of the Software for any purpose unless authorized in writing by GE Healthcare. If authorized to make copies, End-User must mark such copies “COPY” and include a copy of this Software End-User License Agreement . End-User must reproduce proprietary notices on any copies of the Software. End-User is solely responsible to maintain relevant backup procedure and GE Healthcare shall not be liable for any loss of data

1.3.6. Limited WarrantyGE Healthcare warrants that for a period of ninety (90) days from the date of receipt (the “Warranty Period”), that the media on which the Software resides willbe free from defects in materials and workmanship under normal use. In the event that a nonconformity to the foregoing warranty appears during the WarrantyPeriod, End-User must provide GE Healthcare with written notice of the claimed nonconformity. GE Healthcare shall, at its sole option, either: (i) use its commercially reasonable efforts to cure said nonconformity within a reasonableperiod of time; or (ii) replace End-User’s copy of the Software with another copy ofSoftware; or (iii) refund the fees End-User has paid to license the Software. This shall be End-User’s sole and exclusive remedy.

1.3.7. Disclaimer to other WarrantiesNone of the foregoing warranties shall apply if: (i) End-User’s computer malfunctioned and the malfunction caused the nonconformity; or (ii) any other cause within End-User’s control caused the malfunction. GE HEALTHCARE DISCLAIMS ALL OTHER WARRANTIES, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, IMPLIED WARRANTIES FOR MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, WITH RESPECT TO THE SOFTWARE AND ANY ACCOMPANYING WRITTEN MATERIALS. THIS LIMITED WARRANTY GIVES END-USER SPECIFIC LEGAL RIGHTS. END-USER MAY HAVE OTHERS, WHICH VARY FROM STATE TO STATE AND COUNTRY TO COUNTRY. No agent , employee, or representative ofGE Healthcare has any authority to bind GE Healthcare to any affirmation, representation, or warranty concerning the Software; and any affirmation, representation, or warranty made by any agent , employee, or representative shall not be enforceable by End-User.

1.3.8. Damage LiabilityIN NO EVENT WILL GE HEALTHCARE OR ITS SUPPLIERS BE LIABLE (WHETHER IN TORT OR CONTRACT INCLUDING BREACH OF WARRANTY) FOR ANY DAMAGES WHATSOEVER (INCLUDING, WITHOUT LIMITATION, DAMAGES FOR LOSS OF BUSINESS PROFITS, BUSINESS INTERRUPTION, LOSS OF BUSINESS, OR OTHER PECUNIARY LOSS) ARISING FROM THE USE OF OR INABILITY TO USE THESOFTWARE, EVEN IF GE HEALTHCARE HAS BEEN ADVISED OF THE POSSIBILITIES OF SUCH DAMAGES.

28-9035-38UM Chapter 1, Rev A, 2006

7

1.3.9. General ProvisionsThe limitations of liability and ownership rights of GE Healthcare contained herein and End-User’s obligations following termination of this Agreement shall survivethe termination of this Agreement for any reason. End-User may not sublicense, assign, share, pledge, rent or transfer any of its rights under this Agreement in relation to the Software or any portion thereof including documentation. GE Healthcare may assign this agreement to any third party in its absolutediscretion.

No amendments or modifications may be made to this Agreement except in writing signed by both parties.

If one or more provisions of this Agreement are found to be invalid or unenforceable, this Agreement shall not be rendered inoperative but the remaining provisions shall continue in full force and effect .

The Agreement shall be governed by and construed in accordance with the substantive laws of the country where GE Healthcare’s principal or registeredoffice is situated and the parties hereby submit to the non-exclusive jurisdiction of the courts of that country.

6 January 2004

28-9035-38UM Chapter 1, Rev A, 2006

8

2. Introduction to the Multi Target Analysis Module

828-9035-38UM Chapter 2, Rev A, 2006

2.1. Overview

The Multi Target Analysis module is a versatile tool, designed for the analysis of cell based assays where the classification of cells into multiple subpopulations is required. It is applicable to a wide range of assays that utilise fluorescent markers—including cell viability, cell cycle monitoring, apoptosis and cell signalling—but can be applied to almost any assay that examines a cellular process.

Cells can be classified into sub-populations by applying one or more filters, each of which is based on up to two user-selectable fluorescence or morphological measures. For each filter, a histogram or an interactive 2-D scatter plot is available to assist in setting thresholds and discrimination functions. More complex classification schemes can be constructed by combining multiple filters to form a decision tree. Each decision point within the tree allows classification into two populations. A decision tree can therefore be constructed to report multiple subpopulations and also their accompanying measures, thereby allowing subpopulation analysis.

2.2. Example assays and image data

2.2.1. Cell viability

The fluorescent dyes calcein AM and propidium iodide are widely used in combination to assess cell viability. Endogenous esterases active only in living cells hydrolyze calcein AM to the membrane-impermeant green fluorescent product calcein, which is retained within viable cells. In contrast, propidium iodide, a DNA intercalator, is actively excluded from viable cells, and is therefore a marker for nuclei of necrotic and dead cells. The dye, Hoechst 33342 stains the nuclei of both living and dead cells, and therefore can be used as a marker for the total cell population. In this example, human lung epithelial cells (A549) were incubated with increasing concentrations of the toxic compound ionomycin for 24 hours at 37 °C. Fresh medium containing Hoechst, calcein AM and propidium iodide was then added and cells incubated for 10 minutes.

The 2D Linear Discriminant scatter plot filter within the Multi Target Analysis module was configured to identify and measure the green fluorescence of calcein versus the red fluorescence of propidium iodide. Live and dead cell populations can therefore be visualised for classification (Figure 2-1) and the number and percentage of each population reported (Figure 2-2).

928-9035-38UM Chapter 2, Rev A, 2006

Figure 2-1 Cell viability analysis using a Linear Discriminant 2D filter. Cells are identified by segmentation of nuclei in the nuclear marker channel (channel 1). Live cells are identified by the presence of calcein in channel 2. Dead cells are identified by the presence of propidium iodide in the reference 1 channel (channel 3). Slope and intercept of the linear threshold are adjusted manually by the user on a 2D interactive scatter plot to discriminate live and dead subpopulations and the results are displayed in the corresponding bitmap and plate map.

Figure 2-2 Cell viability assay dose response. A549 cells were exposed to increasing concentrations of ionomycin for 24 hours (n=8). Cells were classified according to green and red fluorescence intensities using a Linear Discriminant 2D scatter plot. Non-linear sigmoidal dose-response curves were generated for the concentration ranges shown (mean ± SD) and an EC50 of 17.65 μM ionomycin for cell viability was calculated.

1028-9035-38UM Chapter 2 Rev A, 2006

Figure 2-3 Cell viability and apoptosis analysis using a Decision tree. Cells are identified by segmentation of nuclei in the nuclei marker channel (channel 1). Apoptotic cells are identified by the presence of FITC-Annexin V in channel 2. Dead cells are identified by the presence of propidium iodide in the reference 1 channel (channel 3). Viable cells are identified by the absence of both FITC- Annexin V and propidium iodide. A Decision tree consisting of Threshold and Linear Discriminant 2D filters is generated, with individual linear thresholds adjusted manually by the user to classifiy apoptotic, viable and dead cell subpopulations. Results are displayed in the corresponding bitmap and plate map.

2.2.2. Cell viability and apoptosis

A hallmark of apoptotic and necrotic cell death is the translocation of phosphatidylserine (PS) to the outer leaflet of the plasma membrane. The exposure of PS can be detected using a fluorescently labelled Annexin V conjugate. Use of Annexin V conjugate in combination with viability/non-viability markers such as calcein AM and propidium iodide (section 2.2.1) enables discrimination of apoptotic, viable and dead cell populations.

Depending on the cell type and treatment conditions, Annexin V localizes to phospholipid patches that may appear anywhere within the boundaries of the cell body. The localised fluorescence intensity of the Annexin V and additional markers can be identified and measured by applying 1D and/or 2D filters, which can be combined in a Decision tree to segregate the various sub-populations.

Following a 4 hour incubation at 37 °C with a toxic agent such as ionomycin, cells were incubated for 10 minutes with fresh medium containing FITC- Annexin V conjugate, propidium iodide and Hoechst 33342 nuclear stain. Apoptotic, live and dead cell populations are classified by the Decision tree process and the number and percentage of each population reported (Figure 2-3 and 2-4).

11

Figure 2-4 U-2 OS cells were exposed to increasing concentrations of ionomycin for 4 hours (n=8). Non-linear sigmoidal dose-response curves were generated for the concentration ranges shown (mean ± SD) and an EC50 of 10.92 μM ionomycin for the % Dead population was calculated.

2.2.3. Cell cycle

Cell cycle phase markers are non-perturbing fluorescent proteins that provide an indication of the cell cycle status of individual cells in an asynchronous population. They can be used in live cells with kinetic imaging systems to provide a dynamic, non-invasive method of monitoring the cell cycle, or in end-point format (live or fixed) with other fluorescent probes to highlight complex cell cycle related events.

The G1S Cell Cycle Phase Marker (G1S CCPM, GE Healthcare, Product Code 25-9003-97) exploits functional elements from the human helicase B gene (phosphorylation-dependent sub-cellular localization domain, PSLD) and the human ubiquitin C promoter to serve as a non-perturbing sensor of Cdk2/cyclin E activity. For a given cell in an asynchronous population, the sub-cellular location and intensity of the sensor provide an indication of its cell cycle phase. Incorporation of bromodeoxyuridine (BrdU), which occurs only during S-phase, can be monitored in conjunction with the G1S CCPM reporter to refine cell cycle phase determination.

In this example, U-2 OS cells stably expressing G1S CCPM were treated with increasing concentrations of the cdk inhibitor Roscovitine to block S-phase progression. After 24 hours incubation with Roscovitine, cells were incubated for a further hour with BrdU (Cell Proliferation Fluorescence assay, GE Healthcare, Product code 25-9001-89) to identify cells undergoing DNA replication, and cells were then fixed. BrdU was detected using anti-BrdU primary antibody together with a Cy5-conjugated secondary antibody. Nuclei were stained with Hoechst 33342.

Images were analyzed using the Multi Target Analysis module. A Decision tree was constructed to classify each individual cell into G1, S, G2 or M phase on the basis of sub-cellular location and intensity of the fluorescent reporters (Figure 2-5). The number and percentage of cells identified in each sub-population were reported (see Figure 2-6 for Roscovitine dose response curve). Measures for each of the defined subpopulations can also be reported enabling more detailed sub-population analysis.

28-9035-38UM Chapter 2 Rev A, 2006

12

Figure 2-6 Roscovitine dose response curve (24 hrs) for G1S CCPM fixed cell assay multiplexed with Cell Proliferation Assay imaged on the IN Cell Analyzer 1000 and analysed using the Multi Target Analysis module. An EC50 of 50 μM Roscovitine (% of cells in G1 phase) was calculated from the dose response curve. Mean +/- SD. n=8 replicates per dose. R2 = 0.81.

Figure 2-5 Cell cycle analysis using a Decision tree process. Cells identified by segmentation of nuclei in the nuclear marker channel based on Hoechst 33342 staining (channel 1). G1S CCPM sensor reported using Nucleus / Cell intensity ratio measure to aid classification of G1 and G2 phase cells (channel 2), BrdU incorporation and subsequent immunofluorescent (Cy5) detection facilitates the definition of S phase in the reference 1 channel (channel 3). A Decision tree consisting of Threshold filters is generated with individual threshold values adjusted manually by the user to classify G1, S, G2 and M phase subpopulations. Results are displayed in the corresponding bitmap and plate map.

28-9035-38UM Chapter 2 Rev A, 2006

1328-9035-38UM Chapter 3 Rev A, 2006

3. Workflow Overview and Tips3.1. Assay

• Ensure that the culture plate type is compatible with the instrument . Unless a plate protocol already exists, details about the plate dimensions will be required during acquisition protocol set-up.

• During assay development , optimize seeding density and incubation conditions to ensure that cells will be sub-confluent on the day of image acquisition. (During analysis, object segmentation is most accurate when cells are well- separated; avoid assay conditions that lead to tightly packed cells or cell ‘clumps’).

3.2. Acquisition

• Prior to image acquisition, check that the instrument lamp is properly aligned. In particular, inspect a sample image to check that illumination is even across the sample and that background signal is acceptably low. Contact technical support in case of poor lamp alignment .

• Acquire images from the assay plates. Use a control plate to optimize acquisition parameters. Take care that the signal from features of interest is not saturating (i.e. well within the dynamic range of the camera). During assay development, repeated rounds of acquisition and analysis may be required to optimize acquisition parameters. Typically, fluorescence intensity readings in nuclear and cytoplasmic compartments should be between 30 relative fluorescent units (RFU) above peak background signal and 3500 RFU absolute.

• Ensure that the particular features you want to quantify are in focus and clearly resolved. Increasing the objective power may help to improve definition and contrast of sub-cellular features.

• After acquisition, check several images from different locations across the plate to ensure that image quality is consistent across the plate.

3.3. Analysis

• Create an analysis protocol as described in chapters 4 and 5. You may find batch queue analysis useful during the protocol optimization process (Chapter 10).

• Set up analysis parameters, sample an image or images and apply the appropriate classification filters (Chapter 6).

• Edit the analysis protocol using control wells and various sample wells.

• Analyze the image stack with the newly set analysis protocol, and assess the results. Repeat the process of editing and assessment as necessary until results are satisfactory.

14

4. Initiating the Analysis ApplicationOpening the IN Cell Analyzer 1000 Workstation software launches a start-up wizard that will begin navigating you through the analysis set-up procedure. The following sections describe the steps you follow as you progress through the wizard.

4.1. Selecting a Mode

Select a working mode by choosing one of the following options (Figure 4-1), and then click OK:

• Assay development mode—Select this option if you want to define a new analysis protocol or edit an existing one. If this is the first time you are using the Multi Target Analysis module, this option is recommended.

• Analysis mode—Choose this option if you want to use an existing analysis protocol. You will then be able to analyze an image stack with a predefined analysis protocol, but you will not be able to edit the protocol or define a new one.

If your system administrator has password-protected the workstation software, you will need to enter the password in the indicated field before clicking OK. See your system administrator if you do not have the password.

Note: If you decide you want to change work modes after you have completed the start-up wizard, select the desired mode from the Mode menu in the toolbar located at the top of the IN Cell Analyzer 1000 Workstation window.

Figure 4-1 Mode Selection window.

28-9035-38UM Chapter 4 Rev A, 2006

1528-9035-38UM Chapter 4, Rev A, 2006

4.2. Opening an Image Stack

Clicking OK after mode selection (section 4.1.) launches the window shown in Figure 4-2. Select one of the following options, and then click OK.

• View/analyze image stack—allows you to view the image stack and set up an analysis protocol. This window can also be accessed by choosing Application from the main toolbar and selecting Image stack and analysis, or by clicking the Image stack and analysis icon, found on the main IN Cell Analyzer 1000 Workstation toolbar. This is the preferred choice if you are using the Multi Target Analysis module for the first time

• Open a data file—retrieves images and associated data from a previous analysis

Choosing the first option (above) opens the Image Stack and Analysis window (Figure 4-3), which allows you to open a specific image file, or start a batch queue analysis as described in Chapter 10.

Choose View/Analyze image stack and click OK

Use the resulting browser (Figure 4-4) to navigate to the folder containing the image stack file that you want to analyze.

Figure 4-2 Opening an image stack during start-up. Use this window to view and/or analyze an image stack, or to open a previously analyzed image stack along with its associated analysis data.

Figure 4-3 The Image Stack and Analysis window. Use this window to specify whether you want to view/analyze an image stack or schedule automated batch analysis.

1628-9035-38UM Chapter 4, Rev A, 2006

4.3. Navigating an Image Stack

When you open an image stack, the first images in the stack are displayed along with the Image Stack window (Figure 4-5). Access the appropriate functionalities within the following areas of the Image Stack window (Figure 4-5):

• Preview—On the plate map, click on any well to view its associated image(s). Select wells you want to analyze with test protocols by clicking and dragging the mouse across the wells of interest . Wells to be analyzed will then appear highlighted on the plate map.

• Image—View a well by selecting it from the Well drop-down menu, or toggle through the image stack by clicking + or –. If you have collected multiple images from a well, the Field (field of view) and/or Time (time point) menus will become active, allowing you to access all of the images. The λ menu allows you to select which data channels (wavelengths) are displayed on screen. The Autocontrast function automatically adjusts contrast of the selected image relative to the brightest and dimmest pixels in the image.

• Protocol—The name of the currently selected analysis protocol is displayed in the Protocol area. If you want to change or modify the protocol, click Select to access the Select Protocol window and proceed to section 4.4. Click Analyze to analyze the plate with the currently displayed analysis protocol or right click on a particular well or a set of highlighted wells and select Run analysis on selected images. Click View thumbnails to display thumbnail images of all selected wells.

Figure 4-4 Open Image Stack File window. Use the menu to navigate to the desired image stack file.

1728-9035-38UM Chapter 4, Rev A, 2006

4.4. Using the Select Protocol Window

Use the Select Protocol window to start creating a new analysis protocol, or to edit an existing one.

Click Select on the Image Stack window to show the Select Protocol window(Figure 4-6). Click New… to start the Analysis Protocol Wizard that enables you to develop a new protocol (see chapter 5), or select a protocol from the Protocol name menu, and click Edit… to modify an existing protocol (see Chapter 9).

Figure 4-5 Image Stack window. Use this window to view images in the stack, select and edit protocols, and analyze selected images.

Figure 4-6 Select Protocol Window can be used to modify an existing protocol or to begin a new one.

1828-9035-38UM Chapter 5, Rev A, 2006

5. Creating an Analysis ProtocolThe Analysis Protocol Manager is a starting point from which to launch either theAnalysis Protocol Wizard, which helps you create new analysis protocols, or the Analysis Protocol Editor, which helps you edit existing analysis protocols. This chapter shows you how to progress from the Analysis Protocol Manager through the Analysis Protocol Wizard. (See chapter 9 for instructions on how to use the Analysis Protocol Editor).

Note that when creating a new analysis protocol, you cannot test run the protocoluntil you have completed the Analysis Protocol Wizard. For this reason, you may find it helpful to progress fairly quickly through the Analysis Protocol Wizard, entering approximate values for the various analysis parameters. Once you have created and saved the new protocol, you can switch to analysis mode, analyze a few sample wells, assess the results, and re-open the protocol for further editing.

When the IN Cell Analyzer 1000 software is open, you can launch the analysis protocol manager by choosing Analysis Protocol Manager from the Application menu. Use Next and Back to navigate through the different pages.

5.1. Entering a Protocol Name

When the Analysis Protocol Manager is launched, the first window that appears is the Protocol name window (Figure 5-1). The most recently used protocol will automatically appear in the Protocol name field. Decide whether to Edit , Import , Export , Delete, Rename or create a New protocol.

Figure 5-1 Analysis Protocol Manager window.

To begin creating a new analysis protocol:

1. Click New to launch the Analysis Protocol Wizard.

2. You will be asked whether to use the currently selected protocol as a template(Figure 5-2). Click No.

1928-9035-38UM Chapter 5, Rev A, 2006

3. Type a protocol name into the entry field (Figure 5-3). Choose a unique name forthe protocol (one that does not appear in the Protocol name menu). Note that , at this point , choosing a pre-existing protocol name from the drop-down menu will result in an error message.

4. Click Next.

5. Proceed to section 5.2.

Figure 5-2 Define New Analysis Protocol window.

Figure 5-3 Specifying a name for a new protocol within the Analysis Protocol Wizard. Type a name in the Protocol name field, and any additional information in the Description field (optional).

5.2. Password-Protecting a New AnalysisProtocol

IN Cell Analyzer 1000 software allows assay developers to assign passwords to analysis protocols to prevent unauthorized alterations (Figure 5-4). Password- protection of analysis protocols is optional. If you want to password-protect anew protocol, enter the same password in both the Password and Verify password fields, and click Next. Note that passwords are case-sensitive.If a password is selected, users will not be able to modify the protocol without entering the assigned password.

If you choose not to password-protect your new protocol, leave the two fields blank, and simply click Next to skip to the next step.

2028-9035-38UM Chapter 5, Rev A, 2006

Figure 5-4 Protocol Password window within the Analysis Protocol Wizard. Password-protect a new protocol using this window.

5.3. Specifying the Assay Name andMicroscopy Type

The next step in creating an Analysis Protocol is to select the analysis module that will serve as the basis for the new analysis protocol from the Assay name list in the Assay window (Figure 5-5). To create a new analysis protocol using the Multi Target Analysis module, choose Multi Target Analysis from the Assay name menu. Specify the Microscopy type used in the current experiment (Phase contrast, Fluorescence, DIC, or Bright field). If you are analyzing images acquired from a fluorescently labeled sample, set the Microscopy type to Fluorescence. Note that the options available for Microscopy type will depend on which IN Cell Analyzer system you are using (see the IN Cell Analyzer 1000 instrument user’s guide for more information).

Click Next to proceed to the Multi Target Analysis: Images window (Section 5.4).

2128-9035-38UM Chapter 5, Rev A, 2006

5.4. Defining Input Images

In the Multi Target Analysis: Images window (Figure 5-6), you select which objects you want to measure, and indicate which source image (Wave) contains those objects. Select Object types by clicking in the adjacent check box until a check mark appears. Specify a Wave for each selected Object type by clicking the corresponding Wave entry to reveal a drop-down menu, and then selecting the appropriate source image (Wave 1, Wave 2, Wave 3 or Wave 4).

For Multi Target Analysis, the software pre-selects Nuclei as a required Object type, since the analysis routine requires information from the nuclei compartment. If you are not sure which Wave assignment to make for a particular object type, check the header information in the Image View window (section 4-3) for the image in question. The excitation/emission filter information in the header will help you remember which fluorescent dye is associated with each image channel. The channel number in the header corresponds to the Wave with the same number (i.e. Ch 1 indicates Wave 1, Ch 2 indicates Wave 2, etc.).

The Object types available for selection in Multi Target Analysis include Cells, Organelles and Reference. The Reference Object type can be used to obtain intensity measurements from a specified region from the source image (Wave) of your choice (as explained in section 5.5.5). For example, if you acquired an image of nuclei stained with both Hoechst in Wave 1 and propidium iodide in Wave 3, and wanted to measure the nuclear intensity of both fluorescent stains then you would select Wave 1 as the source for Nuclei (Nuclei is pre-selected as a required Object type) and Reference 1 or Reference 2 as an additional object type with Wave 3 as the corresponding Wave. Later, during segmentation (section 5.5.5), you would indicate that information from the Nuclei channel be used to segment the Reference channel image.

When you have selected Object types and completed the Wave assignments, clickNext to proceed to the Multi Target Analysis: Segmentation window (section 5.5).

Figure 5-5 Analysis Protocol Wizard: Assay window.

2228-9035-38UM Chapter 5, Rev A, 2006

Figure 5-6 The Analysis Protocol Wizard, Multi Target Analysis: Images window. Each object type is associated with an image channel by assigning a corresponding wave number. In the example shown here, (A) the image in channel 1 (Wave 1) will be interrogated for nuclear measurements, while the image in channel 2 (Wave 2) will be interrogated for cell body measurements. (B) The image in channel 3 (Wave 3) will be interrogated for reference 1 measurements.

A

B

2328-9035-38UM Chapter 5, Rev A, 2006

5.5. Using the Multi Target Analysis: Segmentation Window

The next step in defining an analysis protocol is to specify the parameters that will be used for image segmentation. Segmentation is the process of dividing an image into a number of individual objects or contiguous regions, differentiating them from each other and from the image background. When using the Multi Target Analysis: Segmentation window (Figure 5-8), work with each Feature in turn, first selecting a segmentation method (if a choice is available), and then setting the associated segmentation parameters. You access the parameters associated with each feature by clicking the feature name (Nuclei, Cells, Organelles or Reference) from the displayed list.

For each object type, the software offers one or more segmentation methods. Inorder to get meaningful data from an analysis routine, it is important to take the time to determine the correct segmentation method for each type of object , and to optimize the associated segmentation parameters. In most cases, the optimal settings are achieved empirically by systematically testing segmentation methods and parameters.

When setting parameters within the Multi Target Analysis: Segmentation window, you may wish to obtain key measurements directly from representative images. Section 5.5.1. describes the tools available for making these measurements. Sections 5.5.2. to 5.5.5. describe the segmentation methods and parameters available for each object type

5.5.1. Setting Segmentation Parameters on the Basis of Measurements from an Image.

Various context-sensitive icons on the Image Tools toolbar (Figure 5-7) automatically become available when you reach the Segmentation window of the Analysis Protocol Wizard (Figure 5-8). The Image Tools are used to obtain various assay-specific spatial measurements (such as nuclear and cell area) that are used to establish thresholds for object detection. If you have selected Multi Target Analysis as the assay type, the following tools become active, depending on which segmentation method you are working with:

Length: linear—Measures the straight-line distance between two points and returns the length in microns. To make a measurement , select Length: linear, click the pointer in the appropriate measurement field (e.g. Min. granule size), and drag the pointer across the area you want to measure. The measure automatically appears in the entry field. If you repeat this process, each new measurement will overwrite the previous one.

Circle—Allows you to draw a circle or oval with dimensions similar to the object you want to measure, and then returns the area of the shape in square microns. This tool is useful for quickly approximating the size of a typical nucleus or cell body. To use the tool, click on the image that you want to take measurements from, and click on the Circle tool to activate it . Place the pointer in the corresponding measurement field (e.g. Minimum area). The pointer will now appear as circle on the active image. Position the circle over the object you want to measure. Press Ctrl and simultaneously drag the mouse until the pointer shape approximates to the size of the object to be measured. Release Ctrl, center the pointer over the object of interest , and then click the mouse to take the measurement . The area of the resulting circle in square microns is reported in the measurement field. If you repeat this process, each new measurement will overwrite the previous one. Pressing F4 will clear all drawn objects.

24

Figure 5-8 Analysis Protocol Wizard. Multi Target Analysis: Segmentation windows defining Nuclei features (Top-hat).

28-9035-38UM Chapter 5, Rev A, 2006

Outline—Allows you to draw a freehand outline around the representative feature of interest , and returns the enclosed area in square microns. To use this tool, click on the Outline icon, and use the resulting arrow to draw around the feature of interest on the selected image. Click within the outline and the area within the outline will automatically be entered in the measurement field. If you repeat this process, each new measurement will overwrite the previous one. Pressing F4 will clear all drawn objects.

Arrow–Segments an object based on its local background intensity, and returns the measured object’s area in square microns. This tool is particularly useful for estimating the size of irregularly shaped objects. To use this tool, click in the relevant measurement field, click on the image that you want to take measurements from, and click on the Arrow tool. Place the open arrow pointer over the background area immediately adjacent to the object you want to measure. A second, filled arrow will appear. Click on the object you want to measure. An outline appears to indicate the measurement region, and the area of the measured object is returned in the Minimum area field. If you repeat this process, each new measurement will overwrite the previous one. Pressing F4 will clear all drawn objects.

Figure 5-7 Image Tools toolbar. The function of the symbols from left to right are Length: linear, Length trace, Circle, Outline and Arrow. Only the Circle, Outline and Arrow tools are functional when Multi Target Analysis has been selected as the assay type

5.5.2. Nuclei Segmentation

There are three methods of segmentation available for nuclei: Top-hat (See section 5.5.3), Global threshold and Region growing (See section 5.5.4.).

If you use Top Hat segmentation, first enter a value greater than 1.25 in the Minimum area field (Figure 5-8), or use one of the available Image Tools (Circle, Outline, or Arrow; section 5.5.1.) to obtain a minimum area measurement from a representative image. In general, aim to enter a minimum area value that is slightly smaller than the smallest valid nucleus you see. The analysis software will automatically filter out any nuclei having areas smaller than the value you enter.

Note: In the Nuclei segmentation window, the Circle Image Tool is automatically selected from the Image Tools Bar.

2528-9035-38UM Chapter 5, Rev A, 2006

Place the cursor with the circular shape over the target or object of interest. If you wish to change the circle circumference, press down the Ctrl key, then move the mouse cursor without clicking the mouse. The Minimum Area size will be filled in automatically after you click with mouse.

Next set the sensitivity of nuclei detection by adjusting the sliding bar from 0-100 %, or by typing a value in the adjacent field. The sensitivity setting determines which pixel clusters will qualify as objects (nuclei) based on their intensity relative to local background. The lower the sensitivity setting, the brighter a cluster of pixels must be in order to be differentiated from the background; detection of dimmer objects will increase as the percent sensitivity increases.

Select the Filter button (Figure 5-8) to access the object filtering parameters. The resulting Object Filtering window (Figure 5-9) allows objects to be filtered according to their area.

Tick the box to activate this feature. You can choose to keep objects with an area above, below or within a range of values. Enter the required threshold value (or values) into the appropriate fields to include the qualifying objects in the analysis.

Typically Top Hat segmentation is used for identification of nuclei. However, in some cases, you may choose to use Global Threshold (Figure 5-10) or Region growing (Figure 5-11) segmentation methods.

Global threshold segmentation is based on intensity measurements and can be used when nuclei are of varying sizes. This method works best on high contrast images without any background shading. It is not advisable to use Global Threshold segmentation on images with uneven background intensity.

To apply Global threshold segmentation, enter a Threshold value or choose Automatic thresholding by selecting the Automatic checkbox (automatically calculates a minimum intensity above background value for each image). Enter a Minimum Area value in the Minimum area field or use the Circle Image Tool as described above.

Select the Filter button to access the object filtering parameters (described above, and shown in Figure 5-9).

Figure 5-9 Analysis Protocol Wizard. Object Filtering parameters window.

2628-9035-38UM Chapter 5, Rev A, 2006

Figure 5-10 Analysis Protocol Wizard. Multi Target Analysis: Segmentation window defining Nuclei features (Global Threshold).

Region Growing segmentation establishes the nuclei boundaries by incrementally dilating an initial region called the ‘seed image’ until the background (relatively dim pixels) is encountered. This method can be used if the nuclei are of widely differing shapes or sizes.

To use Region Growing segmentation, enter a value in the Minimum area field (Figure 5-11), or use one of the image tools (section 5.5.1.) to obtain measurements directly from a representative image. Only the Circle, Outline, and Arrow tools are available with Region growing. Then choose one of five levels of correction (None, Light, Standard, Enhanced or Heavy) for shading and noise.

Select the Filter button to access the object filtering parameters (described above, and shown in Figure 5-9).

Section 5.5.3. describes the application of this segmentation method.

Figure 5-11 Analysis Protocol Wizard. Multi Target Analysis: Segmentation window defining Nuclei features (Region growing).

When segmentation parameters for Nuclei have been entered, click on the next feature (Cells, Organelles or Reference) that you want to define, and proceed to section 5.5.3., 5.5.4., or 5.5.5. as appropriate.

2728-9035-38UM Chapter 5, Rev A, 2006

5.5.3. Cells Segmentation

From the segmentation menu, choose one of four available methods for cell segmentation:

Top-hat—a relatively rapid transformation that is used to accentuate objects of a specified size, increasing the efficiency of their detection. Top-hat transformation is useful for distinguishing objects from a surrounding uneven background. Choose this method if the objects (cell bodies) you want to identify are fairly uniform in size and shape. Enter the size criterion in the Minimum area field (Figure 5-12), or use one of the image tools (section 5.5.1.) to take measurements from a representative image. Only the Circle, Outline, and Arrow tools are available with Top-hat . Adjust the sensitivity of detection by moving the Sensitivity slider bar or by entering a value from 0-100% in the adjacent field. The sensitivity setting determines which pixel clusters will qualify as objects based on their intensity relative to local background. The lower the sensitivity setting, the brighter a cluster of pixels must be in order to be differentiated from the background; detection of dimmer objects will increase as the percent sensitivity increases.

Select the Filter button to access the object filtering parameters (described in section 5.5.2).

Figure 5-12 Analysis Protocol Wizard. Multi Target Analysis: Segmentation windows defining Cell features (Top-hat).

Multiscale top-hat—provides a means of identifying cell bodies that vary in size. The Cells Multiscale top-hat transform works similarly to that described for Organelles (see Section 5.5.4.). However, for cells segmentation, rather than specifying a size range, you provide a characteristic cell body area; the software will then automatically calculate two detection scales for top-hat transformation. In the Characteristic area field (Figure 5-13), specify an area representative of a typical cell body. Alternatively, use Circle, Outline, or Arrow tools to take measurements from a range of representative cells, and then enter in a mean value based on those measurements. Then adjust the sensitivity of detection by moving the Sensitivity slider bar or by entering a value from 0-100% in the adjacent field. The sensitivity setting functions as described for Top-hat , above.

Select the Filter button to access the object filtering parameters (described in section 5.5.2.).

2828-9035-38UM Chapter 5, Rev A, 2006

Figure 5-13 Analysis Protocol Wizard. Multi Target Analysis: Segmentation windows defining Cell features (Multiscale top-hat).

Region growing—establishes the cell boundaries by incrementally dilating an initial region called the ‘seed image’ until the background (relatively dim pixels) is encountered. The seed image is defined by applying a histogram threshold method, and therefore the seed image comprises the brightest areas in the image.

Use this method if the definition of the cell boundaries is critical for your assay, particularly in cases where shading is minimal and cells are significantly clumped together. Enter a value in the Minimum area field (Figure 5-14), or use one of the image tools (section 5.5.1.) to obtain measurements directly from a representative image. Only the Circle, Outline, and Arrow tools are available with Region growing. Then choose one of five levels of correction (None, Light , Standard, Enhanced or Heavy) for shading and noise. If the background intensity is uneven across the image (Figure 5-15), then shading correction may be necessary. Note that the application of extensive shading correction may distort the shape of cells belonging to cell ‘clumps’. As a result , peripheral areas of the cell bodies may not be detected. If the cell cytoplasm is relatively dim compared to the background, you may need to increase the amount of noise removal (Figure 5-16). Keep in mind that changing the level of Shading removal or Noise removal does not affect the image display. Consequently, to assess the effects of these corrections, you will need to run an analysis and look at the bitmap overlays. If levels of Shading removal and/or Noise removal are too high, legitimate objects may not be outlined by the bitmap overlays. If not enough correction has been applied, too many dim pixel clusters (i.e. areas belonging to the background) maybe outlined by the bitmap overlays. If you are not sure how much correction to apply, incrementally adjust the shading and noise removal levels until the bitmaps correspond as closely as possible to the objects that you want to identify. If the outlines are still not accurate, consider trying one of the other segmentation methods.

Select the Filter button to access the object filtering parameters (described in section 5.5.2.).

2928-9035-38UM Chapter 5, Rev A, 2006

Figure 5-14 Analysis Protocol Wizard. Multi Target Analysis: Segmentation window defining Cells feature (Region growing)

Figure 5-15 Example of shading correction with less Shading removal (left) and more Shading removal (right) applied. Use shading correction if the background intensity is uneven across the image. Note that you will not be able to see the effect of shading correction in the Image View window.

Figure 5-16 An example image before and after shading and noise correction. Use noise correction if the cytoplasmic signal is weak relative to the surrounding background Intensity. Note that you will not be able to see the effect of noise correction in the Image View window.

30

Collar—establishes a ring-shaped cytoplasmic sampling region by dilating outwards a defined distance from the established nuclear region. Use this method if you need to sample cytoplasmic intensity rapidly (for example, in a kinetic assay). Specify the value for dilation by entering a number in the Radius field (Figure 5-17), or use the Length: linear tool to obtain measurements directly from a representative image

Figure 5-17 Analysis Protocol Wizard. Multi Target Analysis: Segmentation window defining Cells feature (Collar).

When segmentation parameters for Cells have been entered, click on the next feature (Organelles or Reference) that you want to define, and proceed to section 5.5.4. or 5.5.5. as appropriate.

5.5.4. Organelles Segmentation

Intracellular Organelles are defined as inclusions for the purposes of detection and segmentaion. The method of segmentation for Organelles is preset to Multiscale top-hat (Figure 5-18). The Multiscale top-hat method provides an efficient means of identifying objects that are heterogeneous in size and shape.

Figure 5-18 Analysis Protocol Wizard. Multi Target Analysis: Segmentation windows defining Organelles features (Multiscale Top-hat)

28-9035-38UM Chapter 5, Rev A, 2006

3128-9035-38UM Chapter 5, Rev A, 2006

Enter any non-zero value into the Min. Granule size field and enter a larger value into the Max. Granule size field. These values specify the lower and upper limits of a size range within which qualifying inclusions must fall. The values you enter should correspond to the minimum cross-section (in microns) of representative inclusions. For guidance in setting realistic values for the granule size limits, use the Length: linear Image Tool (section 5.5.1.) to make measurements from a range of representative objects in sample images.

Next, select from the Scales menu the number of scales (from one to ten) to be used in the segmentation calculation. The number of scales you specify determines how many top-hat transformations the software will apply to the image. Each top-hat transformation will be tuned to a different size within the minimum-maximum range. For example, if you choose three scales, the software will perform three top-hat transformations: one tuned to the minimum specified size, a second tuned to an intermediate size, and a third tuned to the maximum specified size. After performing the specified number of top-hat transformations, the software will combine the information from all of the transformations. The net result will be efficient detection of objects of several different sizes within the specified range. Select a number from the drop-down list . If inclusions fall into one size range, then a scale setting of 1 may suffice (the scale will be based on the value entered in Min size field).

Finally, adjust the sensitivity of inclusion detection using the slider bar, or by entering a value from 0-100% in the adjacent field. The sensitivity setting determines which pixel clusters will qualify as objects (inclusions) based on their intensity relative to local background. The lower the sensitivity setting, the brighter a cluster of pixels must be in order to be differentiated from the background; detection of dimmer objects will increase as the percent sensitivity increases.

The Detect Inclusions field allows you to select the area where inclusions will be detected.

To specify in which regions of the image you wish the software to detect (and quantify) inclusions, choose one of the following options:

• Everywhere— detects all inclusions in any part of the image.

• In the nuclei— detects inclusions in the nuclear region only.

• In the ext . cytoplasm— detects inclusions in the extended cytoplasmic region defined as <dilated cells minus nuclei>.

• In the cells— detects inclusions in the entire cell area

• In the cytoplasm— detects inclusions in the area defined as <cells minus nuclei>

• Out of the cells— detects inclusions anywhere outside the cells, but no organelles within cell regions

• Dilated cells— detects inclusions in the dilated cell region. (Dilated cell outlines are not shown on the bitmap display.)

Any intracellular object can be specified as an inclusion for the purpose of analysis. Specifiying the correct location in which to detect inclusions is therefore an important consideration.

3228-9035-38UM Chapter 5, Rev A, 2006

Figure 5-19 Analysis Protocol Wizard. Advanced segmentation parameters window for Organelle segmentation.

The Advanced segmentation parameters window contains the following options

• Detect inclusion by shape—not enabled in the current software release.

• Sensitivity range—enter a value (from 1.3 to 100) to adjust the contrast interval used to map the inclusions sensitivity scale. The higher the local contrast in the image (specifically, inclusion to background intensity ratio), the higher the sensitivity range value required. The Sensitivity range is set to 4 by default. For most applications, you will not need to change the default setting.

• Smart masking—activate the check box to enable inclusion of inclusion objects that partially overlap with the detection area by more than 50%. For example, if you selected In the cells for the detection area, and more than 50% of the inclusion area overlaps with the cell bitmap, the inclusion is considered to belong to this cell. If less than 50% of the inclusion area overlaps with the cell bitmap, then the inclusion is excluded from the analysis.

Once you have adjusted advanced segmentation parameters, click OK and return to the Multi Target Analysis: Segmentation window.

When segmentation parameters for Organelles have been entered, click on the Reference feature (if required) and proceed to section 5.5.5.

Select the Filter button to access the object filtering parameters (described in section 5.5.2.).

Select the Advanced button to access Advanced Segmentation Parameters(Figure 5-19).

3328-9035-38UM Chapter 5, Rev A, 2006

5.5.5. Reference Segmentation

The method of segmentation for Reference is preset to Pseudo (Figure 5.20). This method takes a copy of the object mask previously generated for Nuclei, Cells or Organelles and applies it to the channel specified as the Source Wave for the Reference. You can choose to use objects from Nuclei, Cells or Organelles by selecting the appropriate option from the drop down menu. For example if a nuclear dye has been used in addition to Hoechst, then this can be sampled by selecting Nuclei from the drop down menu in the Reference: Segmentation window. Two Reference features (Reference 1 and Reference 2) are available in the Multi Target Analysis module.

Figure 5-20 Analysis Protocol Wizard. Multi Target Analysis: Segmentation window defining Reference feature

When segmentation methods have been selected and parameters set, click Next to proceed.

3428-9035-38UM Chapter 5, Rev A, 2006

5.6. Using the Multi Target Analysis: Measures Window

In the Multi Target Analysis: Measures window (Figure 5-21), you specify which measures will be acquired from each image. In the Available measures field, select the check boxes corresponding to the measures you want to acquire. Selected measures are then displayed in the Selected measures field. Each of the available measures is described in Table 5.1. To remove individual measures from the Selected measures list , click the corresponding checkbox again so that the check mark disappears. Alternatively, click Clear selection to clear the entire selection. If you intend to create a new filter (section 5.7.) using one of the available measures, you need to select that measure in the Multi Target Analysis: Measures window first .

Figure 5-21 Analysis Protocol Wizard: Multi Target Analysis Measures window. Measures that have been selected with a check mark appear in the Selected measures list .

After you have made your selections, click Next to proceed.

3528-9035-38UM Chapter 5, Rev A, 2006

Table 5-1 Measures available in the Multi Target Analysis: Measures window. The term ‘intensity’ refers to pixel gray levels. The definitions correspond to individual cell measurements that will be reported in Cell by Cell mode (section 8.4.). When reported in Summary mode, the measurements will be averaged over the population of cells detected in the field of view.

Nuclei Refers to individual cell measurements made in the Nuclei channel

Nuc/Cell Intensity Ratio of intensities sampled in the nuclear and cytoplasm regions

Nuc Area Area of identified nucleus

Nuc cg X X coordinate of nucleus’s center of gravity

Nuc cg Y Y coordinate of nucleus’s center of gravity

Nuc Elongation Mean ratio of the short axis of the nucleus to the long axis of the nucleus. If the value is 1 then the object is center- symmetric (not elongated)

Nuc 1/(Form Factor) Mean nucleus roundness index. Value ranges from 1 to infinity, where 1 is a perfect circle

Nuc Displacement Nuclear displacement is the distance between the nucleus’s and the cell’s centers of gravity, normalized by the gyration radius of the nucleus

Nuc Intensity Average nuclear intensity

Nuc Intensity CV Coefficient of variation of pixel intensities over the population of pixels comprising the nuclear region. In Summary mode (section 8.4.), the reported value is the average of the CVs from all nuclei in the field of view

Cell Intensity Average pixel intensity in the cytoplasm region (within the Nuclei channel)

Compactness Compactness is another measure characterizing shape. It is calculated by the formula: 2*PI*(gyration radius*gyration radius2)/area; Gyration radius is an average radius of a shape

Light Flux (relative) The normalized amount of light emitted by the whole nuclei. It is equal to nucleus average intensity multiplied by area and normalised by cytoplasm average intensity. Nucleus area is taken in pixels. Column title in Summary tables of results is Light Flux

Chord ratio Shortest chord to longest chord ratio. Both chords pass through the shape’s center of gravity

Intensity (N+C) Average intensity in whole cell (nucleus + cytoplasm) (Nuclei channel)

Integrated Intensity The amount of light emitted by the nucleus. It is equal to (Nuc) nucleus average intensity (Nuclei channel) multiplied by nucleus area. Column title in Summary tables of results is IxA (Nuc)

Integrated Intensity The amount of light emitted by the cytoplasm. It is (Cell) equal to cytoplasm average intensity (Nuclei channel) multiplied by cytoplasm area. Column title in Summary tables of results is IxA (Cell)

Integrated Intensity The amount of light emitted by the whole cell. It is equal (Whole Cell) to cell average intensity (Nuclei channel) multiplied by cell area. Column title in Summary tables of results is IxA (N+C)

3628-9035-38UM Chapter 5, Rev A, 2006

Nuclei coordination Refers to the spatial coordinate measurements based on adjacency rules for the determination of whether objects are considered neighbors or not

Spacing (SOI) Measure of the inter-nuclear distance (distance from the center of gravity to the neighbor’s center of gravity, averaged by all neighboring nuclei), determined using the Sphere Of Influence (SOI) adjacency rule

Neighbor count (SOI) Number of neighbouring nuclei, determined using the Sphere Of Influence (SOI) adjacency rule.

Spacing (MIN) Measure of the inter-nuclear distance, determined using the MIN rule

Neighbor count (MIN) Number of neighbouring nuclei, determined using the MIN rule

Spacing (Gabriel) Measure of the inter-nuclear distance, determined using the Gabriel rule

Neighbor count Number of neighbouring nuclei, determined using the (Gabriel) Gabriel rule

Spacing (Lune) Measure of the inter-nuclear distance, determined using the Lune rule

Neighbor count (Lune) Number of neighbouring nuclei, determined using the Lune rule

Cells Refers to individual cell measurements made in the Cells channel

Nuc/Cell Intensity Ratio of nuclear to cytoplasmic intensity values (both values obtained from the Cells channel)

Cell/Bckg Intensity Ratio of cytoplasm to local background intensity values in the Cells channel

Cell Area Cell area

Nuc/Cell Area Ratio of nucleus to cell area

Nuc Intensity Average intensity of pixels within the nuclear region (in the Cells channel)

Nuc Intensity CV Coefficient of variation of pixel intensities over the population of pixels in the nuclear region within the Cells channel. In Summary mode (section 8.4.), the reported value is the average of CVs from all nuclei in the field of view

Cell Gyration Average radius of the shapeRadius

Cell Intensity Average intensity of pixels within the cytoplasm region (in the Cells channel)

Cell Elongation Mean ratio of the short axis of the cell to the long axis of the cell. If the value is 1 then the object is center-symmetric (not elongated). Short and long axes are orthogonal and do not necessarily pass through the shape center of gravity

Cell 1/(Form Factor) Mean cell roundness index. Value ranges from 1 to infinity, where 1 is a perfect circle

3728-9035-38UM Chapter 5, Rev A, 2006

Intensity Spreading Indicator of intensity distribution within the object (cell). If the value is 1, then the intensity is distributed uniformly across the object (cell). If the value is > 1, the intensity concentrates near the periphery of the object . If the value is < 1, the intensity concentrates near the center of the object

Cell cg X X coordinate of cell’s center of gravity

Cell cg Y Y coordinate of cell’s center of gravity

Cell Intensity CV Coefficient of variation of pixel intensities over the population of pixels in the cytoplasm region (within the Cells channel). In Summary mode (section 8.4.), the reported value is the average of CVs from all cells in the field of view.

Background Average intensity (Cells channel) of pixels in the Intensity background immediately adjacent to the cell. The background mask is obtained by dilation of the cytoplasmic outline (not displayed).

Light Flux (relative) The normalized amount of light emitted by the whole nuclei. It is equal to nucleus average intensity multiplied by area and normalized by cytoplasm average intensity (Cells channel). Nucleus area is taken in pixels. Column title in summary tables of results is Light flux.

Intensity (N+C) Average intensity in whole cell (nucleus + cytoplasm)

Integrated Intensity The amount of light emitted by the nucleus. It is equal (Nuc) to nucleus average intensity (Cells channel) multiplied by nucleus area. Column title in Summary tables of results is IxA (Nuc)

Integrated Intensity The amount of light emitted by the cytoplasm. It is equal (Cell) to cytoplasm average intensity (Cells channel) multiplied by cytoplasm area. Column title in Summary tables of results is IxA (Cell)

Integrated Intensity The amount of light emitted by the whole cell. It is equal (Whole Cell) to cell average intensity (Cells channel) multiplied by cell area. Column title in Summary tables of results is IxA (N+C)

Organelles Refers to individual cell measurements of inclusions made in the Organelles channel

Count Number of inclusions attributed to the cell

Spacing Measure of the inter-inclusion distance, determined using the Sphere Of Influence adjacency rule (averaged by all inclusions within the cell)

Neighbor Count Number of neighboring inclusions, determined using the Sphere Of Influence adjacency rule (averaged by all inclusions within the cell)

Mean Area Mean area of inclusions (averaged by all inclusions within the cell)

1/(Form Factor) Mean inclusion roundness index (averaged by all inclusions within the cell). Value ranges from 1 to infinity, where 1 is a perfect circle.

Elongation Mean ratio of the short axis of the inclusion to the long axis of the inclusion (averaged by all inclusions within the cell). If the value is 1 then the object is center-symmetric (not elongated).

3828-9035-38UM Chapter 5, Rev A, 2006

Distance to Nuc Mean distance from the inclusion center of gravity to the nucleus center of gravity (averaged by all inclusions within the cell)

Inclusion/Cell Intensity Ratio of the average pixel intensity within inclusions to the average pixel intensity in the cell region immediately adjacent to the inclusions (both intensities measured in the Inclusion channel)

Intensity Average intensity of pixels within inclusions

Total Area Total area of inclusions attributed to the cell

Inclusion/Bckg Inclusion to background intensity ratio. Background Intensity intensity is measured immediately next to the cell. (Both Inclusion and Background values are obtained in the Inclusion channel)

Reference measures refer to individual cell measurements made in the Nuclei, Cells or Organelles channels, depending on the object selected for Reference segmentation.

5.7. Using the Multi Target Analysis: Filters Window

The Multi Target Analysis: Filters Window (Figure 5.22.) allows you to add threshold filters that identify (1) cells having object measures above or below threshold values, or (2) cells having objects that fall within a range of values. The Multi Target Analysis module also allows you to develop classification protocols using classification filters. Classification filters can only be added when working in Analysis Protocol Editor. (See section 6 for information about how to use threshold and classification filters.)

Note that once a filter has been defined, you will need to select it in the Multi Target Analysis: Summary window (section 5.8.) in order to activate it . Each filter you define will appear on the Multi Target Analysis: Summary window (Figure 5-24), and its cell selection criteria (section 5.9.1) must be defined in order for the appropriate values to be reported.

Figure 5-22 Analysis Protocol Wizard. Multi Target Analysis: Filters window. A filter based on Nuc 1/(Form factor) measure was added (above filter type).

3928-9035-38UM Chapter 5, Rev A, 2006

To add a threshold filter from within the Analysis Protocol Wizard, click Add… . When the window of the available measures appears, highlight the required measure, and then click Select. Remove measures by selecting the measure and clicking Remove….

Once the measure to be used as a filter has been selected, choose a filter type from the Filter type menu. Enter the required information in the context sensitive fields that appear when you select one of the following filter types:

above or below—Enter the threshold above or below which the measurement must be must be in order for the cell to be filtered.

in range—Enter the lower and upper limits within which the measurement mustfall in order for the cell to be filtered.

Clicking calls up the measures list for the highlighted filter, allowing you to edit the filter. Selecting a new measure overwrites the original filter measure, but you will need to manually edit the title to reflect any changes. Each filter you define will appear on the Multi Target Analysis: Summary window (Fig 5.24.) and will be available in the Filter out drop-down menu (see section 5.9.1). Once you have defined all the filters you require, click Next to proceed.

5.8. Using the Multi Target Analysis: Define Classifiers Window

The Multi Target Analysis module allows you to develop analysis protocols that automatically assign cells to pre-defined classes. The protocol used for automatic classification is called a classifier (sometimes also referred to as a classification protocol). Use the Define classifiers window (Figure 5.23) to add any number of classifiers to the analysis protocol. (See chapter 7 for information about how to create a classification protocol.)

Figure 5-23 Analysis Protocol Wizard. Multi Target Analysis: Define classifiers window.

4028-9035-38UM Chapter 5, Rev A, 2006

To add a classifier, click Add… , and then browse to locate the desired classification protocol file. Classifiers can be removed by highlighting the classifier for removal and clicking Remove. Click on the Classification Method, Categories, and Feature Space tabs to find summary information about selected classifiers. Once you have added the desired classifiers, click Next to proceed to theMulti Target Analysis: Summary window.

5.9. Using the Multi Target Analysis: Summary Window

5.9.1 Summary data

Use the Multi Target Analysis: Summary window (Figure 5-24) to specify which measures and classified populations will be included in the data summary output files.

Figure 5-24 Analysis Protocol Wizard: Multi Target Analysis Summary window

The Cells selection area of the window allows you to specify the use of any previously defined filters. In the first drop down menu, you can choose to ‘Include all cells’ in the analysis. Note that if no filters have yet been defined then this is the only option available. If a filter has been defined, then the options to ‘Include only cells where’ or to ‘Exclude cells where’ become available. If either is selected then a further two drop-down menus will be activated (Figure 5-25). The first allows you to choose the required filter on which to base the filter out command. Once the required filter is selected, the second drop down list displays only the populations related to that filter and you can choose to include or exclude cells that pass the filter (True) or cells that fail the filter (False).

Note that an analysis protocol can have only one active Filter out option.

Check the tick box to remove rejected cells from the data set if required.

4128-9035-38UM Chapter 5, Rev A, 2006

Figure 5-25 Analysis Protocol Wizard: Multi Target Analysis Summary window showing filter out options.

In the Measures selection area of the window, select all of the measures and filters for which information output is required by clicking the corresponding check box in the Available Measures list . To remove a measure or filter from the summary list, click the box again so that the check mark disappears. Clicking Clear selection removes all measures from the Summary list. Note that if you change the Reference segmentation method after all parameters have been selected, you will need to re-select the measures and filters associated with that Reference channel (in both the Multi Target Analysis: Measures and the Multi Target Analysis: Filters windows) in order to make them available for selection in the Multi Target Analysis: Summary and Subpopulations windows. Click Next to proceed.

5.9.2. Subpopulation data

Use the Multi Target Analysis: Subpopulations window (Figure 5-26) to specify which subpopulations measures will be reported in the subpopulation data output files (optional and only available if a filter has been defined).

Figure 5-26 Analysis Protocol Wizard: Multi Target Analysis Subpopulations window.

4228-9035-38UM Chapter 5, Rev A, 2006

In the definition of subpopulations area, check the tick box to activate a drop down menu of the filters available for subpopulation definition. Select the required filter from the list and then select the measures required to be reported for each subpopulation, from the available measures list.

Click Next to proceed.

5.10. Using the Time Lapse Stack AnalysisWindow

Use the Time Lapse Stack Analysis window (Figure 5.27) to specify what the software will do if images are missing from the time series. For most experiments, the default option is recommended: ‘Analyze the stack as acquired; if the time point is incomplete, do not analyze it .’. When this option is selected, incomplete time points will be excluded from analysis.

In some cases, you may wish to specify alternative time lapse analysis parameters—for example, when images are missing from the time series or a marker compartment degrades over the time course. To do this, first select the wavelength for which you want to specify time points to be used. (For example, if a nuclear image is missing, select the wave index that corresponds to the nuclear image.) Then select one of the following:

First available—at a given location and wavelength, uses the image from the earliest time point . For example, if a nuclear image was taken at time t = 0, it will be used for the analysis of all time points.

Last available—at a given location and wavelength, uses the image at the latest time point . For example, if a nuclear image was taken at times t = 0, t = 1000, andt = 5000, the image taken at t = 5000 will be used for the analysis of all time points.

Same or closest following—uses the acquired image if available. Otherwise, uses the first available image acquired after the time point of the missing image. For example, if a nuclear image was taken at times t = 0, t = 2000, and t = 5000, and not taken at t = 1000, t = 3000, and t = 4000, the following nuclear images will be used for analysis of the given time point:

4328-9035-38UM Chapter 5, Rev A, 2006

At this time point: Use nuclear image from time point:

0 0

1000 2000

2000 2000

3000 5000

4000 5000

5000 5000

Same or closest preceding—uses the acquired image if available. Otherwise, uses the last available image acquired before the time point of the missing image. For example, if a nuclear image was taken at times t = 0, t = 2000, and t = 5000 and not taken at t = 1000, t = 3000, and t = 4000, the following nuclear images will be used for analysis of the given time point:

At this time point: Use nuclear image from time point:

0 0

1000 0

2000 2000

3000 2000

4000 2000

5000 5000

Finally, click Next to proceed to the Z-Stack Analysis window.

Figure 5-27 Analysis Protocol Wizard.Time Lapse Stack Analysis window

4428-9035-38UM Chapter 5, Rev A, 2006

5.11. Z-Stack Analysis Window

Use the Z-stack Analysis window to specify which planes should be analysed. You can choose to analyse all available planes or specify an individual plane. Select the appropriate option by ticking the box alongside the required option (Figure 5-28).

Click Next to proceed to the Data Management window.

Figure 5-28 Analysis Protocol Wizard. Z- Stack Analysis window

5.12. Using the Data Management Window

Use the Data Management window (Figure 5-29) to specify what the software will do with the acquired data. All files will be named according to the analysis protocol name, and will be saved to the same sub-folder as the image stack file (*.xdce). Click in the corresponding check box to activate as many of the following options as required:

4528-9035-38UM Chapter 5, Rev A, 2006

Figure 5-29 Analysis Protocol Wizard. Data Management window

Save data file—saves the raw analysis output data file (*.lg2 format). The *.lg2 file is required if you later want to re-load the analysis results into the IN Cell Analyzer 1000 workstation. It is also required if you want to annotate an image stack.

Export in MS Excel format—saves the analysis results in the form of an Excel spreadsheet (*.xls format). Only the measures you specified in the Multi Target Analysis: Summary window (section 5.9.) will be reported.

Export as comma-delimited text—saves the analysis results in the form of a comma-delimited text file (*.csv). Only the measures you specified in the Multi Target Analysis: Summary window (section 5.9.) will be reported.

Export as tab-delimited text—saves the analysis results in the form of a tab- delimited text file (*.txt). Only the measures you specified in the Multi Target Analysis: Summary window (section 5.9.) will be reported.

Export as XML text—saves the analysis results in extensible mark-up language format file (*.xml).

You can choose to report all of the results (including summary and individual cell data), or just the population summary data for measures specified in the Summary list (section 5.9.). When using the Multi Target Analysis module it is recommended to select Report all results option.

NOTE: To avoid losing analysis results, do not analyze image stacks after opening them directly from a CD-ROM. Instead, copy the image stack to a hard drive before analyzing it .

When you have finished specifying the output formats, click Finish to save your analysis protocol and exit the Analysis Protocol Wizard.

46

6. Filters and Classification The ability to filter out objects based on any measure and the classification of cells into user defined populations are two functions that are achieved using the filters available in the Multi Target Analysis: Filters Window (Figure 6.1). Filters can be defined using any available measures previously selected in the Multi Target Analysis: Measures window. This allows cells to be filtered or classified according to any fluorescence intensity or morphology based measure.

Filters for classification can only be defined using Analysis Protocol Editor, after a basic analysis protocol has been set up using Analysis Protocol Wizard. An initial analysis of sample wells should also be performed to allow the segmentation parameters to be checked and also to provide sample data to allow filter and classification thresholds to be defined.

Note that once a filter has been defined, it will appear on the Multi Target Analysis: Summary window (Figure 6-21). The use of each filter must be defined for the appropriate values to be filtered out or reported (see section 6.4). Note that if the Reference segmentation method is changed during editing, then any previously defined filter based on measures associated with that Reference channel will be removed from the Multi Target Analysis: Filters window. The appropriate Reference measures will need to be selected again in the Multi Target Analysis: Measures window to be available in the Multi Target Analysis: Filters window.

6.1. Classification using a Threshold filter

Use a threshold filter to define (1) cells having object measures above or below threshold values, or (2) cells having object measures that fall within a range of values. Use this filter type to divide populations into two subpopulations based on a single measure.

For each threshold filter you want to define, click Add… and select Threshold from the filters list to display the Threshold filter window (Figure 6-2).

Figure 6-1 Analysis Protocol Editor. Filters: Classification Filters available.

28-9035-38UM Chapter 6, Rev A, 2006

47

Figure 6-2 Analysis Protocol Editor. Threshold Filter window

In the Title field, enter a name for the threshold filter (optional), and then click to display a list of available measures. (Note that only measures selected in the Multi Target Analysis: Measures window will appear on this list). Highlight the required measures in the list, and then click Select. If you previously selected an analyzed well in the Image Stack window before adding the filter, a histogram will appear in the Threshold window (figure 6-3).

Once the measure to be used as a filter has been selected, choose a filter type from the drop-down menu, in the filter area of the threshold window.

Enter the required information in the context sensitive fields that appear when you select one of the following options:

• above or below—enter the threshold above or below which the measurement must be must be in order for the cell to be filtered

• in range—enter the lower and upper limits within which the measurement must fall in order for the cell to be filtered

28-9035-38UM Chapter 6, Rev A, 2006

48

Figure 6-3 Analysis Protocol Editor. Threshold Filter window showing populated histogram data for Nuc/Cell Intensity (Cells).

Once sample wells have been analysed the histogram will be populated (Figure 6-3). The threshold(s) can then be adjusted by clicking and dragging the threshold line(s) to the required position. The resulting value(s) will then be displayed in the Filter area of the Threshold window. Alternatively values can be entered manually in the appropriate field(s) in the Filter area. In the example shown in Figure 6-3, sample wells have been analysed to populate the histogram and the threshold has been set to separate into two populations based on a threshold value of 1.04 for Nuc/Cell Intensity (Cells).

Select the Graph button to access the graph options (Figure 6-4).

Figure 6-4 Threshold filter: Graph options window.

The Threshold Filter Graph window contains the following options:

• Data display range – the range available for the axis can be changed from Automatic (default setting) to Manual by unchecking the tick box. This activates the Min and Max fields in which appropriate values can be entered. This is useful when setting a threshold value but note that the range setting defaults to automatic when the graph options window is next accessed.

• Histogram bins – the bin size for each histogram bar can be changed from Automatic (default setting) to Manual by unchecking the tick box. This activates the bin width field in which a more appropriate value can be entered.

28-9035-38UM Chapter 6, Rev A, 2006

49

Click OK to activate the selected Graph options.

Select the Classes button to access the Class definition window (Figure 6-5).

Figure 6-5 Threshold filter: Class definition window

The Class definition window contains the following options:

• Color - the color of each class can be changed by right-clicking on the color box and selecting a color from the displayed list.

• Class – the class name can be changed by double-clicking on the current name and entering the required name.

• Symbol – the symbol can be changed by double-clicking on the current symbol and entering the required symbol.

• Area – refers to the specified classification regions (cannot be changed).

Ensure that the name and symbol you have entered corresponds to the correct area (1 or 2) shown in the class definition window.

Click OK to return to the Threshold filter window

The Hide Parent button allows you to hide any previously opened window that may obscure the view of cells in the image windows. This may be useful for checking the identity of cells during classification. Once the Hide Parent button has been selected it changes to Show Parent. Clicking on Show Parent will show all windows again.

The defined Threshold filter then appears in the list of filters in the Multi target Analysis: Filters window (Figure 6-6). To change any of the parameters for the Threshold filter, highlight the required filter in the list and select Edit... Once parameters have been edited, click OK to save the new parameters and to close the filter window. Note that once parameters have been edited, the analysis will have to be run again for the classification bitmap to reflect the updated parameters. The use of the classification filter should then be defined (Section 6.4) before the whole plate is analysed.

28-9035-38UM Chapter 6, Rev A, 2006

50

Figure 6-6 Analysis Protocol Editor. Filters: List of defined Classification Filters. Details of the highlighted filter are shown in the Definition area.

6.2. Classification using a Linear Discriminant 2D filter

The Linear Discriminant 2D filter generates a scatter plot of any two available measures, enabling cells to be classified into up to 4 user defined populations. This filter is useful for applications where distinct sub-populations can be discriminated on the basis of two parameters (e.g. an assay where two different fluorescent dyes are used to mark live and dead cells).

For each Linear Discriminant 2D filter you want to define, click Add… and select Linear Discriminant 2D from the filters list (Figure 6-7) to display the Scatter plot window (Figure 6-8).

Figure 6-7 Analysis Protocol Editor. Multi Target Analysis: Classification filters (Linear Discriminant 2D selected)

28-9035-38UM Chapter 6, Rev A, 2006

51

Figure 6-8 Analysis Protocol Editor. Linear Discriminant 2D (scatter plot) window.

Enter a name for the Linear Discriminant (optional) then click on next to the X: field to display a list of available measures that can be specified for the X axis. Highlight the required measure in the list, and then click Select. Repeat for the Y axis, selecting the required measure.

Only measures selected in the Multi Target Analysis: Measures window will appear in the available measures list. Note also that sample wells will need to have been analysed for the scatter plot to be populated once measures are selected (Figure 6-9).

Figure 6-9 Analysis Protocol Editor. Linear Discriminant 2D filter window showing analysed data.

28-9035-38UM Chapter 6, Rev A, 2006

52

Select the Graph button (Figure 6-9) to access the graph options (Figure 6-10).

Figure 6-10 Linear Discriminant Filter: Graph options window.

The Linear Discriminant 2D Graph window contains the following options:

• X and Y transformation – allows the scale of either or both axes to be changed from no transformation (linear) to either Natural log scale or Base-10 log scale. This may be useful to improve the separation between cell populations.

• X and Y range – the range available for each axis can be changed from Automatic (default setting) to Manual by activating the radio button alongside the required option. This activates the Min and Max fields in which appropriate values can be entered.

Click OK to activate the selected Graph options.

Select the Classes button (Figure 6-9) to access the class definitions options (Figure 6-11).

Figure 6-11 Linear Discriminant Filter: Class definition window (option 1 selected).

28-9035-38UM Chapter 6, Rev A, 2006

Use the class definition window to specify the number and arrangement of classification areas required in the classification protocol. The following options are available:

• Option 1 – allows classification into 2 user-defined populations using 1 linear threshold. Click on the first area option, corresponding to 2 areas (Figure 6-11). The selected class option is then outlined in green and the corresponding color, class and symbol options become available. These can be changed to more relevant descriptions (as described in section 6.1).

• Option 2 – allows classification into 3 user-defined populations (including 1 unclassified) using 2 parallel linear thresholds. To activate this option click on the second area option, corresponding to 3 areas (Figure 6-12). The selected class option is then outlined in green and the corresponding color, class and symbol options become available. These can be changed to more relevant descriptions (as described in section 6.1).

53

Figure 6-12 Linear Discriminant Filter: Class definition window (option 2 selected).

• Option 3 – allows classification into 2, 3 or 4 user-defined populations using 2 intersecting linear thresholds. To activate this option, click on the third area option (it is then outlined in green). An additional function for this option, allows the user to specify the number of populations (up to 4) and their position on the resulting scatter plot (for the 2 population option only) by clicking on the arrows located next to the class option, until the required layout is shown (Figure 6-13). The corresponding color, class and symbol options are then available and can be changed to more relevant descriptions as described above.

Figure 6-13 Linear Discriminant Filter: Class definition window (option 3 selected). (A) 2 classes, (B) 3 classes or (C) 4 classes can be defined for the areas specified.

28-9035-38UM Chapter 6, Rev A, 200628-9035-38UM Chapter 6, Rev A, 2006

Select Refresh (Figure 6-9) to reflect any changes in the dot color that are made when the position of the linear threshold is moved. All dots within the same area will therefore have the same color.

Select Reset (Figure 6-9) to reset the position of the linear thresholds back to their original default positions.

54

Once sample wells have been analysed, the scatter plot will be populated with colored dots (Figure 6-9), each representing individual cells from the analysed population. The linear threshold can then be manually moved by clicking and dragging the threshold line to the required position to specify individual classes of cells. Note that due to the way in which the linear threshold functions, the preferred method is to move the threshold to the required position and click refresh before the color, name and symbol options are changed.

To check the identity of a particular cell and therefore which population it should belong to, click on the required dot on the scatter plot window and the corresponding cell will be highlighted in the image and table view. Use the Hide Parent button (section 6-1) to aid the visualisation of cells and classification bitmap in the image window when checking classification. Note also that for the interaction between a dot on the scatter plot and a cell in the image window to function, the Sample button must be highlighted in the Operations field. It is important to try different combinations of measures to achieve the best separation between classes.

Once the threshold line has been adjusted to the required position, click Refresh to ensure all cells within the same area (class) have the same color. Click OK to close the Linear Discriminant filter window and save the parameters. The Linear Discriminant filter then appears in the list of filters in the Multi Target Analysis: Filters window (Figure 6-14).

To change any of the parameters for the Linear Discriminant 2D filter, highlight the required filter in the list and select Edit. Once parameters have been edited, click OK to save the new parameters and to close the filter window. Note that once parameters have been edited, the analysis will have to be run again for the classification bitmap to reflect the updated classification parameters.

Figure 6-14 Analysis Protocol Editor. Filters: List of defined Classification Filters. Details of the highlighted filter are shown in the Definition area.

28-9035-38UM Chapter 6, Rev A, 200628-9035-38UM Chapter 6, Rev A, 2006

After changes have been made, repeat the analysis of sample wells to check the classification results. Click Select on the plate map window to display the Select protocol window and select Edit. Choose Filters from the protocol list and highlight the Linear Discriminant filter required. Select Edit to display the scatter plot and click on individual cells to check the classification result.

Make changes in the classification parameters (if required) and click OK to close the scatter plot window and save any changes. The use of the classification filter should then be defined (section 6.4) before the whole plate is analysed.

55

6.3. Classification using a Decision tree filter The Decision tree filter allows the user to classify cells into multiple populations based on any available measure. The tree is designed to be multilevel and allows the cell population to be divided into two subpopulations at each decision point. The two types of filter described earlier, Threshold and Linear Discriminant 2D, are available at each decision point for classification and can also be used in combination. At each decision point, either one or both populations can then be classified into further subpopulations or can be reported in the Summary data.

For each Decision tree filter you want to define, click Add… and select Decision tree from the filters list (Figure 6-15) to display the Decision tree window (Figure 6-16).

Figure 6-15 Analysis Protocol Editor. Multi Target Analysis: Classification filters (Decision tree selected)

Figure 6-16 Analysis Protocol Editor. Decision tree window.

28-9035-38UM Chapter 6, Rev A, 2006

56

To set up a Decision tree, click and drag the required node type from the right-hand pane to the larger left-hand work area (hover the mouse over each node icon to reveal the node type).

The available nodes are defined as:

• Start node T – threshold node to be used as first node in a decision tree. Output populations are labelled as 0 and 1.

• Start node S – scatter plot node to be used as the first node in a decision tree. Output populations are labelled as A and B.

• Node T – secondary threshold node. Output populations are labelled as 0 and 1

• Node S – secondary scatter plot node. Output populations are labelled as A and B

The first node in a Decision tree must be a start node and can be a threshold or a scatter plot node. Secondary nodes of the required type (either threshold or scatter plot) are then added to build up a Decision tree. Note that a node can be added and defined immediately or all required nodes can be added first and then defined. Note only one start node can be used in each Decision tree.

Once nodes have been added, right click on each node to display the Edit node option (Figure 6-17). When you select Edit node a Threshold or Linear Discriminant 2D scatter plot window (as appropriate) will then be displayed and is used to define the required populations. Each filter is set up to divide the population into two subpopulations based on the required measures. Filters are edited as described in section 6.1 for a Threshold filter and section 6.2 for a Linear Discriminant 2D filter. Note that there are only two available options within the class definition window for a Linear Discriminant 2D filter, when selected within a Decision tree (Figure 6-18).

Figure 6-17 Analysis Protocol Editor. Decision tree window showing scatter plot decision nodes

28-9035-38UM Chapter 6, Rev A, 2006

57

Figure 6-18 Linear Discriminant Filter, Class Defintion window when selected within Decision tree.

Once populations have been defined for the start node, they can then be reported or separated into further subpopulations:

Hover the mouse over the primary node until the cursor arrow changes to a hand. Click-and-drag to draw a connecting arrow from one sub-population of the start node to the required secondary node. A connecting arrow will then appear (Figure 6-19) linking the two nodes. In the example shown (Figure 6-19), population 0 will be reported in the summary data and population 1 will be separated into two further populations (A and B) based on the filter defined by the secondary node.

Figure 6-19 Analysis Protocol Editor. Decision tree window showing Threshold start node and Linear Discriminant 2D secondary node. Population 0 will be reported and population 1 will be further separated into populations A and B.

28-9035-38UM Chapter 6, Rev A, 2006

58

Right click on the secondary nodes in turn to display the threshold or scatter plot filter (as appropriate) to define the required populations as described previously for the start node.

When defining a decision node filter within a Decision tree, it is important to rename and change the symbol of each population so that the correct populations are reported and can be identified in the summary data table. Note that the names and symbols must be unique for each population and that duplication will result in an error message when the Decision tree window is closed.

Click OK on each filter window to save any changes and click OK on the Decision tree window to save any overall changes. The Decision tree filter then appears in the list of filters in the Multi target Analysis: Filters window (Figure 6-20). You can choose to change the title of the Decision tree from ‘New Decision Tree’ (default).

To change any of the parameters for the Decision tree filter, highlight the required filter in the list and select Edit... Click OK to close the Decision tree window and save any changes. The use of the Decision tree filter should then be defined (section 6.4) before the whole plate is analysed.

Figure 6-20 Analysis Protocol Editor. List of defined filters showing a Decision tree filter.

6.4. Defining the use of the Classification filter

The uses of the classification filters added to the protocol need to be defined in the Multi Target Analysis: Summary window (Figure 6-21) accessed by clicking Summary on the Protocol tree. In the Cell selection area you can choose to include all cells, or to include or exclude cells based on one of the defined classification filters. Only one classification filter can be used to specify the populations for inclusion or exclusion. For each classification filter added, new measures will be added to the Available measures list. Scroll to the bottom of the Available measures list, and check any of the classification filters and measurements that you want the protocol to report (i.e. choose %, n or both).

28-9035-38UM Chapter 6, Rev A, 2006

59

Figure 6-21 Analysis Protocol Editor. Multi Target Analysis:Summary window, specifying which classification measures to report.

If you require subpopulations measures to be reported, these will need to be specified in the Multi Target Analysis: Subpopulations definition window. Click Subpopulations, on the Protocol tree (Figure 6-22) and specify which classification filter to use for subpopulations definition. Then select all of the subpopulations measurements that you want the protocol to report.

Click OK to save the finished analysis protocol.

Figure 6-22 Analysis Protocol Editor. Multi Target Analysis: Subpopulations window, specifying which subpopulation measures to report.

28-9035-38UM Chapter 6, Rev A, 2006

60

6.5. Example Classification Protocols

The example protocols provided in this section describe how the classification filters could be used individually or as part of a Decision tree, to analyse three different assays.

6.5.1. Cell viability

This protocol has been designed to analyse a cell viability assay utilising Hoechst 33342 nuclear dye, together with propidium iodide to identify dead cells and calcein AM to identify live cells. Two populations have been classified using a Linear Discriminant 2D filter.

Hoechst 33342 has been used to identify the nuclei of all cells and is required by the Multi Target Analysis module for nuclear segmentation. Wave 1 has been selected to detect Hoechst in channel 1. Cells have been selected as an additional object type with Wave 2 as the source to detect the calcein fluorescence in channel 2. Reference 1 has also been selected as an object type with Wave 3 as the source, to detect propidium iodide in channel 3. For reference segmentation, the method selected uses objects defined by nuclei. Information has therefore been taken from the nuclei segmentation mask using channel 3, to measure the red fluorescence of propidium iodide.

Segmentation parameters were first optimised and sample wells then analysed. A Linear Discriminant 2D scatter plot has then been added. The measures selected to plot were Integrated Intensity of nuclei in the cells channel (IxA (Nuc) (Cells)) for the X-axis, to detect the green fluorescence of calcein (live cells), and Integrated intensity of nuclei in the reference channel (IxA (Nuc) (Ref)) for the Y-axis, to detect the red fluorescence of propidium iodide and hence dead cells. The resulting scatter plot shows two main populations, representing live and dead cells (Figure 6-23). The scale of the axes has been changed to Base-10 log for improved separation between populations.

Figure 6-23 Linear Discriminant 2D scatter plot showing two main populations of cells after initial analysis of cell viability assay sample wells.

28-9035-38UM Chapter 6, Rev A, 2006

61

A region between the live and dead cell populations contains difficult to segment cells, where two cells have been identified as one object and where one cell is live and one dead. These have been removed from the analysis by selecting the three populations area option, corresponding to two classified and one unclassified population. The color, class name and symbol of each population for this option have been changed appropriately (Figure 6-24).

Figure 6-24 Linear Discriminant 2D Class definition window

The threshold has then been manually altered to define each population and the identity of selected cells confirmed by clicking on the required dot on the scatter plot and checking the corresponding cell in the image window (Figure 6-25). Analysis of the sample wells has then been repeated to check the classification parameters (Figure 6-26).

Figure 6-25. Screenshot of interactive Linear Discriminant 2D scatter plot showing identified dot and corresponding cell and bitmap in image window.

28-9035-38UM Chapter 6, Rev A, 2006

62

Figure 6-26 Screenshot of classification of cell viability assay showing control and treated sample wells and corresponding classification bitmap.

In the Summary window, the New Linear Discriminant filter is selected to report the number and percent of live and dead cells. Unclassified cells are selected to be excluded and removed from the analysis (Figure 6-27) and the whole plate then analysed (Figure 6-28).

Figure 6-27. Analysis Protocol Editor: Summary window. Classification is based on the Linear Discriminant filter and the Unclassified cells selected to be excluded and removed from the reported measures.

28-9035-38UM Chapter 6, Rev A, 2006

63

Figure 6-28 Screenshot of classification of cell viability assay showing analysis of whole plate. Summary by fields data is selected showing the percentage Live and Dead results, with % Dead results shown in the color plate map. The classification bitmap for well F11 (highlighted) is shown in the Image view for Channel 2 (Cells).

6.5.2. Apoptosis

This example protocol has been designed to analyse an apoptosis assay where a FITC- labelled Annexin V conjugate has been used to identify early apoptotic cells (based on phosphatidylserine exposure). The Annexin V conjugate has been used together with propidium iodide to discriminate dead cells and Hoechst 33342, which stains the nuclei of all cells. Those cells without the additional fluorescence of either FITC-Annexin V or propidium iodide are discriminated as viable cells. The analysis protocol uses a Decision tree to classifiy cells into one of the above cell populations and reports the number and percentage of each subpopulation, together with additional user-defined measures.

Hoechst 33342 has been used to identify the nuclei of all cells and is required by the Multi Target Analysis module for Nuclei segmentation. Wave 1 has been selected to detect Hoechst in channel 1. Cells have been selected as an additional object type with Wave 2 as the source to detect the FITC- Annexin V conjugate fluorescence signal in channel 2. Reference 1 has also been selected as an object type with Wave 3 as the source, to detect the propidium iodide fluorescence signal in channel 3 (Figure 6-29).

28-9035-38UM Chapter 6, Rev A, 2006

64

Segmentation parameters were optimised. Top-hat segmentation was selected for the Nuclei feature (Minimum area = 50μm2, Sensitivity = 80). Collar segmentation was selected for the Cells feature (Radius = 2μm). Pseudo segmentation using objects defined by nuclei was selected for the Reference 1.

Sample wells were then analysed to check the segmentation parameters and also to provide sample data for the Decision tree filters. A Decision tree using a Threshold node as the start node and a Linear Discriminant 2D scatter plot node as the secondary node was then added to the protocol (Figure 6-30) using Analysis Protocol Editor.

Figure 6-29 Multi Target Analysis Segmentation features and sources for the apoptosis example classification protocol.

Figure 6-30 Apoptosis example assay Decision tree.

28-9035-38UM Chapter 6, Rev A, 2006

65

A start threshold node (Start Node T) was used to filter cells using Nuc Intensity (Reference 1). Any cells with compromised membranes were stained red with propidium iodide fluorescence and therefore had a high Nuc intensity in the Reference 1 channel. Neither viable nor early apoptotic cells exhibited this fluorescence and were therefore discriminated. Using the Cell-by-Cell data of Nuc Intensity (Reference 1) and the threshold histogram (with Nuc Intensity (Reference 1) plotted on the x-axis), a threshold value of 270 for Nuc Intensity (Reference 1) was chosen (Figure 6-31). Cells with a Nuc Intensity (Reference 1) greater than or equal to 270 were classified as Dead cells (Symbol D, Bitmap colour Red) and were then reported. Cells with a Nuc Intensity (Reference 1) less than 270 were classified as ‘not red’ (Symbol NR, Bitmap colour Black) (Figure 6-32). These cells were then taken on to the secondary node for further classification.

Figure 6-31 Start Node T in the apoptosis Decision tree, populated with sample data.

Figure 6-32 Class definition for Start Node.

28-9035-38UM Chapter 6, Rev A, 2006

66

Cells classified as ‘not red’ by the start node, were then further classified using a Linear Discriminant 2D scatter plot node. Cells were discriminated into viable and apoptotic cell populations based on using the Nuc/Cell Intensity (Nuclei) measure on the x-axis and the Cell Intensity (Cells) measure on the y-axis (Figure 6-33). The ratiometric measure Nuc/Cell Intensity was chosen as it provided good separation between the required populations and also compensated for any variation in the fluorescence intensity within a well. The apoptotic cells are discriminated as they have an increased Nuc/Cell Intensity (Nuclei) value compared to the viable cells. Discrimination into apoptotic and viable populations could not have been achieved based on the Cell Intensity (Cells) measure only, due to the increased levels of unbound FITC annexin V fluorescence in the control wells causing viable cells to exhibit an increased Cell Intensity (Cells) value.

The identity of cells was confirmed by clicking on the required dot in the scatter plot and checking the cell in the image window. The linear threshold was then moved to the required position to classify viable and apoptotic cells. The classes were refreshed to ensure all dots in the same population had the correct color, as defined in the Class definition window (Figure 6-34).

Figure 6-33 Secondary scatter plot node discriminating viable and apoptotic populations.

Figure 6-34 Class definition for Linear Discriminant 2D scatter plot node.

28-9035-38UM Chapter 6, Rev A, 2006

67

The entire plate was then analysed, ensuring the ‘include all cells ‘ option and the Decision tree summary measures (% and n) were selected in the Summary window of Analysis Protocol Editor (Figure 6-35). Subpopulation measures were also selected for reporting, based on the Decision tree filter (Figure 6-36).

Figure 6-35 Analysis Protocol Editor, Summary window. Classification includes all cells and the Decision tree measures (% and n) reported for each population.

Figure 6-36 Analysis Protocol Editor, Subpopulation window. Subpopulations were defined using the Decision tree filter and measures selected for reporting.

The results of the whole plate analysis for the example apoptosis classification protocol are shown in Figure 6-37. The cell classification bitmap overlay is highlighted on the images (right) and the plate map with colour coded wells showing % apoptotic cells is also shown (left).

28-9035-38UM Chapter 6, Rev A, 2006

68

Figure 6-37 Screenshot of the Multi Target Analysis module after whole plate analysis. The cell classification bitmap overlay (right) shows apoptotic and dead cells and the plate map with colour coded wells shows the % apoptotic cells.

6.5.3. Cell Cycle

This example protocol has been designed to analyse images generated on the IN Cell Analyzer 1000 of a fixed cell endpoint cell cycle assay. The G1S Cell Cycle Phase Marker (G1S CCPM) (GE Healthcare, Product Code 25-9003-97) cell line are U-2 OS cells stably expressing the G1S CCPM sensor. Cells were treated with increasing concentrations of the cdk inhibitor Roscovitine for 24 hours at 37oC. The Cell Proliferation Fluorescence assay (GE Healthcare, Product code 25-9001-89) was also used for bromodeoxyuridine (BrdU) incorporation in cells undergoing DNA replication (indicated by anti-BrdU monoclonal antibody and Cy-5 immunofluorescence) and Hoechst 33342 was used for nuclear staining. This protocol segments cells, generates a number of measures for each cell and classifies individual cells into one of four populations (G1, S, G2 and M phases) using three threshold node filters within the Decision tree.

Hoechst 33342 has been used to identify the nuclei of all cells and is required by the Multi Target Analysis module for Nuclei feature segmentation. Wave 1 has been selected to detect Hoechst 33342 in Channel 1. Cells feature has been selected as an additional object type with Wave 2 as the source to detect the G1S CCPM sensor fluorescence signal in Channel 2. Reference 1 feature has also been selected as an additional object type with Wave 3 as the source to detect BrdU-Cy5 fluorescence signal in Channel 3 (Figure 6-38).

28-9035-38UM Chapter 6, Rev A, 2006

69

Figure 6-38 Multi Target Analysis Segmentation features and sources for the cell cycle example classification protocol.

Segmentation parameters were optimised. Top-hat segmentation was selected for the Nuclei feature (Minimum area = 80μm2, Sensitivity = 98). Region growing segmentation was selected for the Cells feature (Minimum area = 100μm2, Shading removal – light, Noise removal – light). Pseudo segmentation using objects defined by nuclei was selected for the Reference 1 feature, therefore the Nuclei feature segmentation mask from Channel 1 was used in Channel 3 to measure the BrdU-Cy5 fluorescence signal.

Sample wells were analysed (Figure 6-39) to check the segmentation parameters and also to provide sample data for the Decision tree filters.

Figure 6-39 Plate map with colour coded wells showing Cell Count in sample wells analysed with the optimised segmentation parameters in the cell cycle example classification protocol.

A Decision tree (Title G1S CCPM) using three threshold node filters was then added to the protocol (Figure 6-40 and Figure 6-41).

28-9035-38UM Chapter 6, Rev A, 2006

70

Figure 6-40 Decision tree filter (Title G1S CCPM) added using the Analysis Protocol Editor.

Figure 6-41 G1S CCPM Decision tree in the cell cycle example classification protocol.

28-9035-38UM Chapter 6, Rev A, 2006

71

The threshold node Start Node T (Title T1) was used to filter cells using the Nuc/Cell Intensity (Reference 1) ratio. Those cells that were undergoing DNA replication (cells in S phase of the cell cycle) and therefore incorporated BrdU were labelled using the Cell Proliferation Fluorescence assay and had a higher Nuc/Cell Intensity in the Reference 1 feature than those cells that did not incorporate BrdU. Using the Cell-by-Cell data of Nuc/Cell Intensity (Reference 1) and the threshold histogram (with Nuc/Cell Intensity (Reference 1) plotted on the x-axis) a threshold value of 1.2 Nuc/Cell Intensity (Reference 1) was chosen (Figure 6-42). Cells with a Nuc/Cell Intensity (Reference 1) greater than or equal to 1.2 were classified as S phase cells (Symbol S, Bitmap colour Red) and reported. Cells with a Nuc/Cell Intensity (Reference 1) less than 1.2 were classified as ‘To Be Classified Further’ (Symbol TBCF1, Bitmap colour Black) (Figure 6-43). To further help in accurately setting the threshold value, the data display range of the histogram was manually chosen to be min 0.5 and max 4.5 Nuc/Cell Intensity (Reference 1) and the Histogram bin width chosen to be 0.05 (Figure 6-44).

Figure 6-42 Start Node T (Title T1) in the G1S CCPM Decision tree.

Figure 6-43 Class definition for Start Node T (Title T1) in G1S CCPM Decision tree.

28-9035-38UM Chapter 6, Rev A, 2006

72

Figure 6-44 Histogram definition window for Start Node T (Title T1) in the G1S CCPM Decision tree.

Those cells classified as ‘To Be Classified Further’ by Start Node T (Title T1) were filtered through Node T (Title T2). The threshold node, Node T (Title T2) was used to filter cells using the Nuc/Cell Intensity (Cells) ratio. The phenotype of G2 phase cells shows the G1S CCPM sensor to be mainly expressed in the cytoplasm, with very little expression in the nuclear region. Therefore the Nuc/Cell Intensity (Cells) ratio of G2 phase cells will be lower than G1, S and M phase cells. Using the Cell-by-Cell data of Nuc/Cell Intensity (Cells) and the threshold histogram (with Nuc/Cell Intensity (Cells) plotted on the x-axis) a threshold value of 1.05 Nuc/Cell Intensity (Cells) was chosen (Figure 6-45). Cells with a Nuc/Cell Intensity (Cells) greater than or equal to 1.05 were classified as ‘To Be Classified Further’ (Symbol TBCF2, Bitmap colour Black). Cells with a Nuc/Cell Intensity (Cells) less than 1.05 were classified as G2 phase cells (Symbol G2, Bitmap colour Yellow) and reported (Figure 6-46). To further help in accurately setting the threshold value, the data display range of the histogram was manually chosen to be min 0.8 and max 1.6 Nuc/Cell Intensity (Cells) and the Histogram bin width chosen to be 0.01 (Figure 6-47).

Figure 6-45 Node T (Title T2) in the G1S CCPM Decision tree.

28-9035-38UM Chapter 6, Rev A, 2006

73

Figure 6-46 Class definition for Node T (Title T2) in G1S CCPM Decision tree.

Figure 6-47 Histogram definition window for Node T (Title T2) in the G1S CCPM Decision tree.

Those cells classified as ‘To Be Classified Further’ after Start Node T (Title T2) were filtered through Node T (Title T3). The threshold node, Node T (Title T3) was used to filter cells using the Integrated Intensity IxA (Nuc) measure. In a fixed cell assay under standard, non-saturated imaging conditions, the Integrated Intensity IxA (Nuc) of Hoechst 33342 fluorescence relates to the amount of DNA in the nucleus and is therefore useful for determining the nuclear DNA content of cells. Using the Cell-by-Cell data Integrated Intensity IxA (Nuc) and the threshold histogram (with Integrated Intensity IxA (Nuc) plotted on the x-axis) a threshold value of 140000 Integrated Intensity IxA (Nuc) was chosen (Figure 6-48). Cells with an Integrated Intensity IxA (Nuc) greater than or equal to 140000 were classified as in M phase of the cell cycle (Symbol M, Bitmap colour Blue) and reported. Cells with an Integrated Intensity IxA (Nuc) less than 140000 were classified as in G1 phase of the cell cycle (Symbol G1, Bitmap colour Green) and reported (Figure 6-49). To further help in accurately setting the threshold value, the data display range of the histogram was automatically chosen to be min 0 and max 150000 Integrated Intensity IxA (Nuc) and the Histogram bin width chosen to be 3193.3 (Figure 6-50).

28-9035-38UM Chapter 6, Rev A, 2006

74

Figure 6-48 Node T (Title T3) in the G1S CCPM Decision tree.

Figure 6-49 Class definition for Node T (Title T3) in G1S CCPM Decision tree.

Figure 6-50 Histogram definition window for Node T (Title T3) in the G1S CCPM Decision tree.

28-9035-38UM Chapter 6, Rev A, 2006

75

Therefore using this example Decision tree classification strategy, each cell will be classified into G1, S, G2 or M phase of the cell cycle using the following criteria. Note however, very few M phase cells were observed in this fixed cell assay since they are rounded and poorly attached and are easily removed during fixation, exposure to anti-BrdU antibodies and washing steps.

S phase Nuc/Cell Intensity (Reference 1) greater than or equal to 1.2 (T1).

G2 phase Nuc/Cell Intensity (Reference 1) less than 1.2 (T1) and Nuc/Cell Intensity (Cells) less than 1.05 (T2).

G1 phase Nuc/Cell Intensity (Reference 1) less than 1.2 (T1) and Nuc/Cell Intensity (Cells) greater than or equal to 1.05 (T2) and Integrated Intensity IxA (Nuc) less than 140000 (T3).

M phase Nuc/Cell Intensity (Reference 1) less than 1.2 (T1) and Nuc/Cell Intensity (Cells) greater than or equal to 1.05 (T2) and Integrated Intensity IxA (Nuc) greater than or equal to 140000 (T3).

The entire plate was then analysed, ensuring the Decision tree (G1S CCPM) summary measures (% and n) were selected in the Measure Selection region and including all cells in the Cell Selection region of the Analysis Protocol Editor (Figure 6-51).

Figure 6-51 Analysis Protocol Editor, Summary window. Classification includes all cells and the G1S CCPM Decision tree measures (% and n in each population) are reported.

The results of the example cell cycle classification protocol (including Decision tree) are shown in Figure 6-52. The cell classification bitmap overlays are highlighted on the images (right) and the plate map with colour coded wells showing % of cells in G1 phase is also highlight (left) after analysing the entire plate with the example cell cycle classification protocol. The user should note that the classification scheme represents one of a number of potential strategies that could be employed to analyze the same cell cycle assay using the Multi Target Analysis module.

28-9035-38UM Chapter 6, Rev A, 2006

76

Figure 6-52 Screenshot of the Multi Target Analysis module after analysis of the assay plate with the example cell cycle classification protocol including Decision tree. Cell classification bitmap overlay is shown (right) and the plate map with colour coded wells detailing % of cells in G1 phase is also shown (left).

28-9035-38UM Chapter 6, Rev A, 2006

77

7. Using Supervised Classifiers7.1. Overview

The Multi Target Analysis module has optional supervised classification capabilities that enable you to train the software to assign individual cells to distinct classes. During supervised classification, you can define a number of phenotypically distinct cell classes. This chapter provides instructions for assigning representative cells to defined classes (annotation), and then training the software to recognize those classes (classification). The completed classification routine is then saved as a classification protocol file that can be incorporated into any Multi Target Analysis protocol. Note that the classification routine generated within the supervised classifiers function must be used independently of the classification filters.

7.2. Preparing for annotation and classification

For accurate annotation and classification, it is important to work from the highest quality image data possible. Take care during image acquisition to ensure that the features characteristic of each cell class are in focus and that the signal intensity of the features is within the dynamic range of the camera (typically between 50 and 3500 gray levels (RFU)). Before you begin annotation, review the image data and decide how many classes you need to describe the different types of cells observed in the population. Make a note of what combination of features best describes each class. Then assess whether the image set contains adequate numbers of cells representative of each class. The more cells you annotate to each class, the more reliable the resulting classification protocol will be. It is typical to annotate between 30 and 100 cells for each defined class.

Once you are satisfied that the image set is suitable for annotation, perform an analysis run on the images, taking care to optimize the analysis protocol so that you accurately detect all of the features that you think will be important for classification. Distinguishing features can include shape, size or relative intensity of any cell feature. For example, an unhealthy cell might have intensely stained inclusions (e.g. Annexin V staining), condensed nuclear DNA, and an altered cell body shape

7.3. Using the annotation tool

A prerequisite for annotation and classification is the existence of an analysis data file (*.lg2 format). To open an analysis data file, click Sample in the Operations window (Figure 7-1); choose Open from the File menu, and browse to find the appropriate file. Alternatively, analyze an image set (see section 4.3. and/or section 8); be sure to click Done to complete the analysis. If you have chosen to analyze only a sub-set of the images in the image stack, the software will ask you whether you want to save the data; click OK. Once the analysis data file has been saved, any classifiers you subsequently define using these data will automatically be associated with the corresponding *.lg2 file.

28-9035-38UM Chapter 7, Rev A, 2006

78

Figure 7-1 Operations toolbar showing the Sample and Visuals icons

You use the Annotation Tool to define the cell classes of interest, and then manually assign representative cells to each class. To open the Annotation Tool window, first select Cell by Cell from the data display menu on the Table Toolbar (Figure 7-2), and then click the Annotation icon on the Table Toolbar. The Annotation Tool window will appear (Figure 7-3).

Figure 7-2 The Table Toolbar. The Table Toolbar contains options for displaying and exporting data after an analysis has been completed. Tools, from left to right, are: Data Display Menu, Show targets, Data Export, Annotation, Classification Protocol Builder, and Outline Display Options.

Figure 7-3 Annotation Tool window.

28-9035-38UM Chapter 7, Rev A, 2006

79

The Annotation Tool window is organized into three pages: Classes, Annotation, and Values. When defining a new classification protocol, use the Classes page first, and then the Annotation and Values pages as necessary. The Classes, Annotation and Values pages are described in sections 7.3.1., 7.3.2., and 7.3.3., respectively.

7.3.1. Using the Classes page to define cell classes

Use the Classes page (Figure 7-4) to name and color-code cell classes. View the Classes page by clicking on the Classes tab. Add new classes by right-clicking anywhere within the field (located below the Symbol and Name headers), and then selecting Add from the resulting menu. Repeat this action until you have as many classes as you require (i.e. one class for each cell type you want to categorize). For example, if a typical cell population is composed of three distinct types of cells, create three new classes (Figure 7-4).

Figure 7-4 Annotation Tool window showing Classes

To edit the class details, right-click on each class name to access the following options:

Add—Creates another class.

Remove—Removes the selected class.

Rename—Allows you to assign the selected class a meaningful name. For example, you might choose ‘Apoptotic’ for one class and ‘Healthy’ for another.

Color—Allows you to assign one of ten colors to the selected class.

You can also retrieve and edit previously created class files by clicking Load. When the Retrieve Classes window appears (Figure 7-5), browse to find the desired class file (*.xann), and click Open. To modify the retrieved class, right-click on its name to access the editing options (as above), and then click Update to accept the changes.

28-9035-38UM Chapter 7, Rev A, 2006

80

Figure 7-5 Retrieve classes window

Once you have defined and color-coded the classes, you can save this information by clicking Save and using the resulting Save Classes browser. Then proceed to section 7.3.2. to begin annotation.

7.3.2. Using the Annotation page to assign cells to defined classes

Open the Annotation page by clicking on the Annotation tab within the Annotation Tool window (Figure 7-6). An image of the first cell for classification will be displayed on the Annotation page. (If no image appears, double click an Image View window to select the image.) Use < and > at the bottom of the annotation page to toggle between the available image channels (waves) associated with the current well.

Figure 7-6 Annotation Tool window showing the Annotation page

28-9035-38UM Chapter 7, Rev A, 2006

81

To select a cell for classification, click on individual cells in any of the Image View windows. Alternatively, use < and > at the top of the Annotation page to toggle through each cell in the selected image. To switch from one well to the next, use << and >>. A display at the bottom of the image on the Annotation page indicates the image channel (CH), the degree of magnification (1:×), and the class to which the selected cell has been assigned (if any).

Clear… allows you to clear all previously defined classification for the Selected cell, the entire Current well, or All cells available. Objects cleared of any annotation information will be marked ‘???’ (unknown classification). Cells of unknown classification will not be used by the classification protocol builder (section 7.4.).

To annotate a selected cell, first click on the class to which you think the displayed cell belongs, and then click Annotate. If this is the first time annotation data has been collected from the image stack using the Apoptosis analysis module, the software will tell you that the current data set has not yet been prepared for annotation. Click Yes to proceed.

When making class assignments, you may find it helpful to show the analysis overlay for the cell to verify which cell is being annotated and how the cell features have been segmented. Clicking Show overlays repeatedly toggles the analysis overlay on and off (see Figure 7-7). You may also find it helpful to use information on the Values page (section 7.3.3.) to help you decide to which class a cell belongs (Figure 7-8).

Figure 7-7 Annotation Tool window showing the Annotation page with the bitmap overlays

28-9035-38UM Chapter 7, Rev A, 2006

82

At any point during annotation, the class assignment for a particular cell can be altered by re-selecting the cell, choosing a different class, and then clicking Annotate. The new annotation will over-write the previous one. For each class, the software requires at least ten annotated cells. It is typical to annotate between 30 and 100 cells for each defined class. However, there is no upper limit to the number of cells you can assign to a given class. As a general guideline, the more subtle the differences between the classes, the more cells you should annotate to ensure adequate differentiation.

The file containing all of the annotated data is sometimes referred to as a ‘training data set’ because it will be used to train a classifier used in the classification protocol (section 7.4.). Once you have annotated sufficient cells for each defined class, click Close to save the annotation file, and then proceed to section 7.4. to build a classifier protocol.

7.3.3. Using information in the Values page

To view the Values page (Figure 7-8), click the Values tab within the Annotation Tool window. The Values page displays the analysis measurements for the cell being annotated. Use this information to help you decide to which class a cell should be assigned

Figure 7-8 Annotation Tool window showing Values page

28-9035-38UM Chapter 7, Rev A, 2006

83

7.4. Building a classification protocol

The Classification Protocol Builder uses the training data set you created during annotation to create an automated protocol that can classify cells into defined categories. You can incorporate the finalized protocol into any new Apoptosis analysis module.

Launch the Classification Protocol Builder window (Figure 7-9) by clicking the Classification Protocol Builder icon on the Table Toolbar (Figure 7-2).

Figure 7-9 Classification Protocol Builder window

The classification categories you created during annotation will appear in the Classification Protocol Builder window, along with the number of cells (‘# Patterns’) you annotated for each class.

Click Next to proceed to the Protocol Name window (Figure 7-10), where you can enter a name and a description for the classification protocol. Click Next and proceed to section 7.4.1.

Figure 7-10 Classification ProtocolBuilder: Protocol Name and ProtocolDescription window

28-9035-38UM Chapter 7, Rev A, 2006

84

7.4.1. Choosing a classification algorithm

The analysis software uses a classification algorithm to group cells into defined classes based on information from the training data set. In the Classification Algorithm window (Figure 7-11), select one of the following algorithms:

Nearest neighbours—For each cell it classifies, the nearest neighbours algorithm considers the five nearest classified points in multi-dimensional feature space, and assigns the cell in question to the majority class. (The dimensions of feature space are determined by the features you select in section 7.4.2.) Try this method if the classes have complex phenotypes that are likely to be difficult to distinguish. The nearest neighbours method is non-parametric and can handle situations in which cells comprising the classes in the training data set are not distributed normally.

Neural network—This algorithm examines the training data set to ‘learn’ how to handle complex data. For the learning element of the algorithm to work well, large numbers of cells must be annotated for each class. Try this method if the classes are likely to be difficult to distinguish, and you have annotated a large number of cells for each class. Although this method is parametric, it is able to handle complex data in multi-dimensional space.

Quadratic Discriminant—Try this method if annotated examples of each class are easily differentiated—particularly when each class is homogeneous in appearance. The quadratic discriminant algorithm uses a parametric probabilistic classification model, producing non-linear (second-order) curved boundaries between classes in feature space.

In most cases, the choice of classification algorithm is empirical. If you are not sure which method is likely to work the best, it is advisable to try each of the methods in turn.

When you have selected a classification algorithm, click Next to proceed to feature selection (section 7.4.2.).

Figure 7-11 Classification ProtocolBuilder: Classification Algorithmwindow

28-9035-38UM Chapter 7, Rev A, 2006

85

7.4.2. Selecting features for classification

The window shown in Figure 7-12A. allows you to choose which features will be used to assign cells to the defined classes.

In the Available Features field, click check boxes that correspond to features you want to include. Then click Add Checked >> to add all of checked features to the Selected Features list. Alternatively add any feature (checked or unchecked) by highlighting it , and clicking Add >>. Use Remove (for highlighted individual features) and Remove All to clear features from the Selected Features list.

Figure 7-12 Selecting features forinclusion in a classification protocol.(A) Using the Add Checked >>function, eight items have been added to the Selected Features list. (B) The corresponding CVA Scatter Plot shows how well two cell classes (Blue = inactive, Green = active) are separated in feature space using the eight selected features.

Use of the check box feature in building the Selected Features list is particularly helpful when you are deciding which features are most important for classification. If you have used this method, you will be able to remove one or more features from the Selected Features list, assess the impact using the CVA Scatter Plot (see Figure 7-12B.), and then easily add the features back if required by clicking Add Checked >>.

28-9035-38UM Chapter 7, Rev A, 2006

A

B

86

Keep in mind that the more features selected for the classification algorithm, the more cells you will need to have annotated (section 7.3.2.) in order to produce a robust classification protocol. As a general rule, the number of annotated cells required for each class increases to the power of n, where n is the number of features chosen to define the classes.

Clicking Show CVA Scatterplot initiates canonical variance analysis of the annotated cells using the selected features. The results are collapsed into two dimensions, and graphed on a scatter-plot (Figures 7-12B and 7-13B) that helps you visualize how effectively the different classes of cells comprising the training data set are separated in feature space. At the bottom of the scatter plot, the following metrics are displayed:

Goodness (Metric)—Metric measure of inter-class separation in multi-dimensional feature space, based on inter-class to intra-class distance ratio. The larger the value, the better the class separation in feature space.

Goodness (KS pValue)—Alternative measure of inter-class separation based on applying the non-parametric Kolmogorov-Smirnov statistical test to the data. The larger the value, the better the class separation in feature space

It is advisable to keep the classification protocol as simple as possible by minimizing the number of features required to achieve adequate discrimination of the defined classes. Use the Feature Space Optimization section of the window (Fig 7-12A) to help identify which features are the most important in discriminating cell classes. First, in the Sub-space Dimensionality field, specify the number of features (N < 10) that you want to define the feature space. Then click Find Optimal Sub-Space. The software will calculate which N selected features are the most useful for class discrimination, and display them in the Selected Features list. You can then assess the optimized sub-space by clicking Show CVA Scatterplot. For example, for the protocol being developed in Figure 7-12A, eight features were added to the Selected Measures list. Subsequently, a value of four was chosen for sub-space dimensionality. After clicking Find Optimal Sub-Space, the Selected Features list shows which four of the eight original selected features make the largest contribution to the discrimination of the two cell classes (Figure 7-13A). The associated CVA scatterplot (Figure 7-13B) shows that the optimized four-parameter sub-space provides improved discrimination when compared with the original eight selected measures.

28-9035-38UM Chapter 7, Rev A, 2006

Figure 7-13 Assessing results after optimizing sub-space. (A) After applying the Find Optimal Sub-Space function, only the most influential four (of the original eight) selected features appear in the Selected Features list.

(B) The CVA Scatter Plot shows that optimizing subspace has improved discrimination of the two cell classes and also reduced the number of required features.

A

87

B

A typical strategy for optimizing the Selected Features list is to choose up to nine features that appear to be important for class discrimination, and then use the Find Optimal Sub-Space function to decrease of the number of Selected Features incrementally to the lowest number that allows you to achieve acceptable discrimination. The workflow for this is as follows:

1 Use the Add Checked >> function to add features to the Selected Features list that you think are important for discriminating the different classes.

2 Using the CVA scatter plot, assess how well the classes are discriminated.

3 Reduce the number of selected features to nine by systematically removing features from the list and assessing the results with the CVA scatter plot.

4 In the Available Features list, make sure that only the nine features selected in step 3 are checked.

5 Click Remove All to clear the Selected Features list, and then click Add Checked >> to refresh the list with the desired nine features.

6 Click Show CVA Scatterplot, and make a note of the Goodness measures.

7 Set the Sub-Space Dimensionality (start with a value of eight).

8 Click Find Optimal Sub-Space.

9 Click Show CVA Scatterplot; record the current sub-space dimensionality and corresponding Goodness metrics.

28-9035-38UM Chapter 7, Rev A, 2006

88

10 Click Add Checked >> to add back the features that were removed during optimization.

11 Repeat steps 7-10, each time reducing the Sub-Space Dimensionality by one, until results for all dimensionalities have been recorded.

12 Review the results, and choose the conditions that achieve adequate class discrimination with the minimum number of features

If you cannot achieve adequate discrimination even with a large number of features, you may need to choose different features, select a different classification algorithm, or annotate more cells per class.

When you have optimized the Selected Features list, click Next to proceed.

7.4.3. Saving the Classification Protocol

The final window in the Classification Protocol Builder (Fig 7-14) lets you train the classifier and save the resulting classification protocol to a file (*.xcls format). Click Train and Save… to train the classifier. When prompted, designate a name and folder for the classification protocol, and click OK. When you are done, click Finish to exit the Classification Protocol Builder.

Figure 7-14 Classification ProtocolBuilder: Train and Save… window

28-9035-38UM Chapter 7, Rev A, 2006

89

7.4.4. Adding the Classification to the Analysis Protocol

You can incorporate one or more saved classification protocols into any Apoptosis analysis protocol. In Assay Development mode (section 4.1.), select Analysis Protocol Manager from the Application menu. Click Edit to open the Analysis Protocol Editor (Figure 7-15). On the Protocol tree, within the Analysis Protocol Editor, click Supervised Classifiers, and click Add…. Browse to locate the desired classification protocol, and click Open. The name of the opened protocol will appear in the Analysis Protocol Editor window (Figure 7-15).

If you add more than one classification protocol to an analysis protocol, each classification protocol will be applied sequentially during analysis.

Figure 7-15 Analysis Protocol Editor window. Classification protocols are imported into an analysis protocol using the Add… function. The name of the imported classifier appears in the Define classifiers field. Information about any imported classifiers is located on the Classification Method, Categories and Feature Space tabs.

For each classification protocol imported into an analysis protocol, new measures will be added to the Available measures list within the Apoptosis: Summary window (Figure 7-16). Before saving the analysis protocol, you need to specify which (if any) of these measures you want the analysis protocol to report. To do this, open the Summary page (Figure 7-16) by clicking Summary on the Protocol tree. Scroll to the bottom of the Available Measures list, and check any of the annotation measurements that you want the protocol to report (i.e. choose %, n or both). Click OK to finish incorporation of the classifier and save the finished analysis protocol.

Figure 7-16 Specifying which classification measures the analysis protocol will report within the Apoptosis: Summary window. When a classification protocol is imported into an analysis protocol, related summary measures are added to the Available measures list. These measures (% and n) will not be reported in the summary analysis results unless you activate the respective check boxes before saving the analysis protocol

28-9035-38UM Chapter 7, Rev A, 2006

90

8. Analyzing Images and Assessingthe Results8.1. Analyzing an Image Stack

To analyze an image stack using a defined analysis protocol, you need to access the Image Stack window.

If the IN Cell Analyzer 1000 Workstation software is not already open, launch it and then select one of the following options:

Assay Development—Choose this mode if you want to test an analysis protocol and refine it by making further modifications.

Assay Analysis—Choose this mode if you want to execute analysis using a previously optimized analysis protocol.

Follow the instructions to open the image stack that you want to analyze, and then select an analysis protocol. In the Image Stack window, click Analyze to initiate analysis of the entire plate. Alternatively, if you want to analyze only selected wells, highlight the wells of interest , right click over the selection, and then select Run analysis on selected images.

If the IN Cell Analyzer 1000 Workstation software is already open, select Assay Development or Assay Analysis from the Mode menu, and then follow the instructions to open the image stack that you want to analyze. (Note that the desired image stack may already be open if you have just exited the Analysis Protocol Wizard). In the Image Stack window, click Analyze to initiate analysis of the entire plate. Alternatively, if you want to analyze only selected wells, highlight the wells of interest , right click over the selection, and then select Run analysis on selected images.

The analysis protocol examines each image in the stack, automatically identifies features of interest , and makes all the measurements selected in the analysis protocol. As the analysis proceeds, the color-coding on the plate map in the Image Stack window changes to indicate the current analysis status. White indicates wells for which no image was obtained; green indicates a well that has not been analyzed; pink indicates that a well is currently being analyzed; and a shade of orange indicates wells that have been analyzed.

After images in a well have been analyzed, the plate color-coding for that well changes to provide an indication of the analysis results (see section 8.3. for more information).

Various graphical displays and bitmap overlays are available to help you assess performance of the analysis protocol (sections 8.2. to 8.4.). When you are finished analyzing and assessing the data, click OK. The analysis results you specified during creation of the analysis protocol will be saved in the same folder as the image stack file.

28-9035-38UM Chapter 8, Rev A, 2006

91

8.2. Using Graphical Displays to AssessAnalysis Protocol Performance

When optimizing an analysis protocol, you will need to test protocol performance frequently. Use available visual displays and bitmap overlays to assess the protocol.

To plot the results for analyzed wells, right click on the analyzed wells, and select Plot Selected Data. The results will be displayed as a scatter-plot based on default settings (Figure 8-1).

Figure 8-1 Graph of data

To change this view, click Options…. In the resulting Graph Options window (Figure 8-2) specify the graph type. (Selected Field Only is non-functional in this analysis module). If you have chosen histogram as the graph type, specify the measure you want plotted on the x-axis. If you have chosen to display the data as a scatterplot , specify which measures you want plotted on both the x- and y-axes. Then click Apply. In the example shown in Figure 8-3, the graph type has been changed to histogram, and the Nuc 1/(Form factor) has been plotted on the x-axis. The Colors section of the Graph Options window is non functional in this analysis module.

Figure 8-2 Graph Options window

28-9035-38UM Chapter 8, Rev A, 2006

92

Figure 8-3 Graph Options: Histogram window

8.3. Colour-Coding the Plate Map

Once the analysis of selected wells is complete, the color-coding associated with the plate map changes to reflect the analysis results. You can specify how the plate map is color-coded using the Color Bar Properties window (Figure 8-4). Open the Color Bar Properties window by first clicking on >> in the Image Stack window to reveal the color bar, and then clicking Details, which appears below the color bar.

Select the measure that you want to color-code from the Measure menu, and enter a value (from 3 to 32 inclusive) in the Number of Levels field. The data for the selected measure will be divided into as many equally sized bins as you specify in the Number of Levels field. Each bin will be represented by a different color on the plate map. To automatically adjust the color-coding so that it is scaled to match data range, click the Automatic Range check box. Alternatively, you can manually adjust the color range by entering values in the From and To fields. When you are finished, click Apply to implement your choices, and Close to exit the window.

Figure 8-4 Color Bar Properties window

28-9035-38UM Chapter 8, Rev A, 2006

93

Figure 8-5 The completed analysis

Figure 8-6 Post-analysis screenshot. The bitmap overlay shows the cell classes (in this case, blue indicates viable cells, red indicates dead cells, green indicates apoptotic cells. The plate map color coding has been set to show % Apoptotic results (see section 8.3.).

28-9035-38UM Chapter 8, Rev A, 2006

94

8.4. Viewing Analysis Results within theWorkstation Software

When analysis is complete, the workstation software will appear, similar to the screenshot shown in Figure 8-5. Bitmap outlines indicating object segmentation and any classification results will automatically be superimposed over the gray-scale images in the Image View windows. To superimpose the results bitmap over the merged (multi-color) image, click on the image twice (once to activate the image, and once to apply the bitmap). If your analysis protocol reports classification measures (section 6), the bitmap overlays will indicate the class to which each cell has been assigned (Figure 8-6).

The Data window (Figure 8-8) displays population or cell-by-cell analysis results. To view the Data window, make sure that you have clicked Sample in the Operations window (Figure 8-7).

Figure 8-7 Operations window showing Sample and Visuals options and Table toolbar menu.

Then choose Summary, Subpopulations (if available), or Cell by Cell from the Table Toolbar menu (Figure 8-7) to switch between population summary data (average results for all the cells identified in the image) and cell-by-cell data (data acquired from each identified cell in the image). Use the scroll bars to view all the different measurements that were specified in the analysis protocol. Clicking Clear discards all of the current analysis data. After you clear the data, it will no longer be displayed in the Data window.

28-9035-38UM Chapter 8, Rev A, 2006

95

Figure 8-8 Post analysis Data window with classification summary data for each well.

Click any line of data in the Summary data table to display its corresponding images in the Image View window. Bitmap overlays are automatically superimposed over the corresponding images to allow you to assess how closely the segmentation corresponds to your visual perception of features of interest. While assessing the segmentation, you can click F4 to temporarily clear thebitmap overlay. To permanently clear the bitmap, move to another well by clicking another location on plate map, and then return to the original well. You can then re-apply the bitmap, if desired, using the Pointer tool (below).

You can use the Pixel Info tool on the main menu to gain spatial and intensity information directly from the selected image (Figure 8-9). Note that this toolbar is context sensitive. If you find that the tools are inactive, toggle between the Sample and Visuals icons in the Operations window (Figure 8-7) to activate them. Click on Pixel Info icon to activate the following tools:

Pointer—displays analysis bitmap overlays when you click on the active image.In Summary mode, click anywhere within the image to superimpose bitmaps over all detected cells in the image. In Cell by Cell mode, click on each individual cell to display its bitmap. (If no bitmap appears, the cell was not analyzed.)

Pixel-info—displays the x-y coordinates and intensity (gray-level) of individual pixels. When this tool is activated, cross-hairs will be displayed when you hover the mouse over the active Image View window. Click and drag the mouse over the image to display x-y coordinates and intensity of the pixel at the center of the cross-hairs.

28-9035-38UM Chapter 8, Rev A, 2006

96

Ruler—use the Ruler tool to display a scale bar on the image. When the tool is active, a scale bar is displayed when you hover the mouse over the active Image View window. Click the mouse to deposit a copy of the ruler on the image. To remove all the rulers from a single Image View pane, click F4. Be aware that the Ruler will only be accurate if the magnification changer on the microscope was in the default position (1X) during image acquisition. Click on the Pointer icon to return to normal.

Figure 8-9 Pointer, Pixel Info, andRuler Toolbar

You can adjust the color-coding and appearance of bitmap overlays by clicking on the Outline Display Options icon on the Table Toolbar (Figure 8-10). This opens the Outline Display Options window (See Figure 8-11).

Figure 8-10 The Table Toolbar. The Table Toolbar contains options for displaying and exporting data after an analysis has been completed. Tools, from left to right , are: Data Display Menu, Show Targets, Data Export , Annotation, Classification Protocol Builder Annotation, and Outline Display Options

You can select Cytoplasm Display options from the list of available options.Right clicking on any of the features allows you to change the color of the associated bitmap. You can select to Display cell label option to visualise the results of classification. You can choose the Display cell outline only option to display the object boundaries or de-activate this option to show the entire area of objects detected in the cells channel.

Figure 8-11 Outline Display Options window

8.5. Viewing Previous Analyses

To open a previously analyzed image stack along with its associated analysis data, click File on the main menu and select Open… from the drop-down menu. Use the resulting browser to locate the analysis file (*.lg2) of interest , and then double click on the filename to open it .

To view the analysis results, click Application on the main menu and select Image Stack and Analysis from the drop-down menu. Choose View / analyze current image stack and click OK.

28-9035-38UM Chapter 8, Rev A, 2006

97

9. Using the Analysis Protocol EditorThe Analysis Protocol Editor allows you to review and refine analysis protocols, and is accessible when the workstation is in Assay Development mode. You can open the Analysis Protocol Editor using either the Image Stack and Analysis application or the Analysis Protocol Manager application:

• Section 9.1: Access via Image Stack and Analysis—allows you to navigate the stack using the Image Stack window, edit the analysis parameters, and test the protocol by analyzing selected images using the Sample function. The Image Stack window and the Analysis Protocol Editor can be open concurrently so that segmentation measurements can easily be extracted from the image for incorporation into the analysis protocol, and population data can be generated for editing classification filters. (For information on how to open and use the Image Stack window, see Chapter 4.)

• Section 9.2: Access via Analysis Protocol Manager—allows you to view and edit analysis parameters, and then use the Test function to exit the Protocol Editor and test the protocol on an image stack using the Image Stack window. The Image Stack window and the Analysis Protocol Editor cannot be open concurrently.

Note: If you want to edit classification filters, it is preferable to open Analysis Protocol Editor using the Image Stack and Analysis application (section 9.1) rather than the Analysis Protocol Manager application. In the latter case, you will not be able to access the population distribution plot through the Test function.

9.1. Analysis Protocol Editor accessed through the Image Stack and Analysis application

To open a protocol for editing, check that you are in Assay Development mode, and then select Image Stack and Analysis from the Application menu. In the resulting Image Stack window (figure 9-1), click Select…, to open the Select Protocol window (figure 9-2). Then select the protocol you want to modify using the drop-down menu associated with the Protocol name field (figure 9-2).

28-9035-38UM Chapter 9, Rev A, 2006

98

Figure 9-2 Select Protocol window.

28-9035-38UM Chapter 9, Rev A, 2006

Figure 9-1 Image Stack Viewer window, accessed in Assay Development mode via Image Stack and Analysis (see Chapter 4).

Note that only those assay protocols compatible with the current image stack will be available in the protocol list for editing. For example, if the current image stack has 3 image channels then only those protocols containing 3 Waves or less will appear in the list.

99

Figure 9-3 Analysis Protocol Editor window when accessed with the Image Stack and Analysis application.

Clicking Edit… opens the Analysis Protocol Editor window (Figure 9-3) containing the Protocol tree through which you can access the parameters you want to change.

On the Protocol tree in the left-hand pane, click on any item that you want to edit to access the associated protocol details. For example in Figure 9-3, the user clicked on Filters, to open the Multi Target Analysis: Filters window. In this window, you can edit the classification filters as described in chapter 6.

The Sample function within the Analysis Protocol Editor allows you to analyze selected wells so that you can assess performance of the current segmentation parameters. The Sample function is context-sensitive, and will only be activated when you are editing Segmentation. Click Sample to initiate a sample analysis on wells that you have selected in the Image Stack window. (See section 4.3 to review details on how to use the Image Stack window to navigate the image stack and select wells). Following a sample analysis, image segmentation results will be displayed in the Image View windows.

Note that if you change the Reference segmentation method during editing, you will need to re-select the measures and filters associated with that Reference channel (in both the Multi Target Analysis: Measures and the Multi Target Analysis: Filters windows) in order to make them available for selection in the Multi Target Analysis: Summary and Subpopulations windows.

Once the modifications have been made, click OK on the Analysis Protocol Editor window to save any changes. Once you have exited the protocol editor, you can continue to use the Image Stack window (as described in section 4.3) to view and analyze images of interest using the modified protocol. Assess the analysis results by examining summary data, graphical displays and segmentation outlines. Continue to edit and test the analysis parameters as described above, until you are satisfied with the protocol.

28-9035-38UM Chapter 9, Rev A, 2006

10028-9035-38UM Chapter 9, Rev A, 2006

9.2. Analysis Protocol Editor accessed through Analysis Protocol Manager

To access the Analysis Protocol Manager, check that you are in Assay Development mode, and then select Analysis Protocol Manager from the Applications menu.

From the Analysis Protocol Manager window (Figure 9-4), select the protocol you want to modify from the drop-down menu in the Protocol name field, and then click Edit…. The resulting Analysis Protocol Editor window (Figure 9-5) contains the Protocol tree through which you can access the parameters you want to change. On the Protocol tree in the left-hand pane, click on any item that you want to edit to access the associated protocol details.

Figure 9-4 Analysis Protocol Manager window.

Figure 9-5 Analysis Protocol Editor window when accessed through Analysis Protocol Manager.

10128-9035-38UM Chapter 9, Rev A, 2006

If you change the Reference segmentation method during editing, you will need to re-select the measures and filters associated with that Reference channel (in both the Multi Target Analysis: Measures and the Multi Target Analysis: Filters windows) in order to make them available for selection in the Multi Target Analysis: Summary and Subpopulations windows.

The Test function allows you to temporarily exit the Analysis Protocol Editor and then test-run the modified protocol on a sample image stack using the Image Stack window. Click Test… to exit the protocol editor and open the Image Stack Viewer window. If no compatible image stack is open, the software will prompt you to open one. If the selected image stack does not meet the requirements of the analysis protocol, you will be prompted to open a different image stack. Once an appropriate image stack has been opened, navigate and analyze the image stack as described in section 4.3. Assess the analysis results by examining summary data, graphical displays and segmentation outlines.

Once the analysis is complete, click Done to return to the Analysis Protocol Editor. The software will prompt you to save the current data (optional). Continue to edit and test the analysis parameters until you are satisfied with the modified protocol. Click OK to exit the Analysis Protocol Editor and automatically save the modifications.

Note that because the Analysis Protocol Editor and the Image Stack window cannot be open simultaneously when accessed with the Analysis Protocol Manager, you will not be able to generate and display population data on selected wells when you are editing the classification filters. For this reason, to edit classification filters it is recommended that you open the Analysis Protocol Editor using the Image Stack and Analysis application (section 9.1).

102

10. Using the Batch Queue10.1. Using the Batch Queue

A batch queue is a scheduling protocol for automated analysis of specified image stacks with one or more analysis protocols. Once you initiate analysis by starting a batch queue protocol, no further operator intervention is required, so you can leave the data to be analyzed while you are away from the Workstation. If you have accumulated image stacks from multiple experiments, each of which requires a different analysis protocol, you can easily schedule the analysis to run sequentially.

The batch queue function is also advantageous during analysis protocol optimization, enabling you to analyze the same image stack repeatedly with a series of test protocols.

10.2. Creating the Batch Queue Folder

Before you set up your first batch queue, you need to specify a directory for the batch analysis folder. (In other words, you need to alter the default file path settings.) Select Settings from the main menu and click Default File Paths… In the Default File Path Settings window (Figure 10-1), click in the entry field for the Batch analysis folder, type BATCH at the end of the file path after the last backslash and then click OK. Click Yes, when the software asks whether you want to create the directory.

Figure 10-1 Creating the Batch analysis folder, BATCH, for the initial batch analysis

28-9035-38UM Chapter 10, Rev A, 2006

103

10.3. Opening the Batch Queue Manager

Selecting Batch Queue Manager from the Workstation Application menu opens the Batch Queue Manager (Figure 10-2). At the same time, an analysis queue file named queue.txt is automatically created in the batch analysis folder(Figure 10-3).

Figure 10-2 Batch Queue Manager window

Figure 10-3 Automatic generation of the file queue.txt in the batch queue folder, BATCH

28-9035-38UM Chapter 10, Rev A, 2006

104

10.4. Adding Image Stacks and Analysis Protocols to the Batch Queue

In the Batch Queue Manager window (Figure 10-2), click Add… to open the Add Image Stacks to Batch Queue window (Figure 10-4). Then click Select ... to open the Select Protocol window (Figure 10-5A). Choose an analysis protocol from the drop-down menu. Click View details… to review the analysis parameters (Figure 10-5B). Note, however, that while it is possible to make apparent changes to the protocol from the Analysis Protocol Viewer, the changes will be lost when you click Done. If you need to edit the analysis protocol, return to the Analysis Protocol Editor and make changes as described in Chapter 9. Once you have chosen an analysis protocol, click OK.

Next , in the Add Image Stacks to Batch Queue window (Figure 10-4), click to open a browser that will enable you to navigate to the folder containing the image stack file(s) you want to analyze (Figure 10-6).

Figure 10-4 Add Image Stacks to Batch Queue window.

28-9035-38UM Chapter 10, Rev A, 2006

105

Figure 10-5 Selecting a protocol for batch analysis. (A) The Select Protocol window, where you can select a named protocol from the drop-down menu and view protocol details by clicking View Details; (B) The Analysis Protocol Viewer, where you can review (but not edit) all the parameters for the selected analysis protocol.

A

BFigure 10-6 Browsing for folders containing your image stacks

28-9035-38UM Chapter 10, Rev A, 2006

106

Once you have located the desired folder, click OK to exit the browser. Then click Scan in the Add Image Stacks to Batch Queue window to start scanning the selected folder for image stacks. The scanning options portion of the Add Image Stacks to Batch Queue window has a default specification to search for *.xdce format files when you click Scan (Figure 10-4). However, depending on which version of IN Cell Analyzer Workstation software you are using, you may want to specify additional or alternative search criteria. You can specify up to four types of file formats (*.xdce, *.xde, *.xml, and *.run) by clicking in the corresponding check boxes. If you select Match stacks to analysis protocol, all image stacks of the specified file format will appear in the scan results (Figure 10-7A), but check boxes corresponding to image stacks that are incompatible with the selected protocol will be grayed out so that they cannot be selected. If you select Hide stacks not matching the current protocol, only those image stacks that are compatible with the selected protocol will appear in the scan results (Figure 10-7B).

Figure 10-7 Adding image stacks to a batch queue. (A) After scanning with Match stacks to analysis protocol selected, image stacks incompatible with the current analysis protocol are displayed but inaccessible (grayed check boxes); (B) After scanning with Hide stacks not matching the current protocol selected, only image stacks compatible with the current analysis protocol are displayed; (C) Clicking the corresponding check box selects an image stack for addition to the batch queue.

A

B28-9035-38UM Chapter 10, Rev A, 2006

107

CWhen you have located the image stack files you want to analyze, click on their corresponding check boxes to select the image stack(s) for analysis (Figure 10-7C), and then click OK to add the selected image stack files to the batch queue. If this is the first time a particular analysis protocol has been added to the batch analysis folder, the software will ask you to select a file name for it (Figure 10-8). Click OK, specify a name for the protocol in the Export Analysis Protocols window (Figure 10-9A), and then click Save. In the batch analysis folder, the software will create a new protocols folder (Figure 10-9B) in which all protocol files (*.xanp) for batch analysis will be saved. The software will then return you to the Batch Queue Manager, where you will see the specified image stack file and its associated analysis protocol in the image stack queue (Figure 10-10).

Figure 10-8 Warning about the analysis protocol.

28-9035-38UM Chapter 10, Rev A, 2006

108

Figure 10-9 Adding analysis protocols to the batch queue. (A) Specify a name for a protocol the first time it is added to the batch analysis folder;(B) A folder called ‘protocols’ is automatically created in the batch analysis folder the first time you create a batch queue. All protocols for batch analysis will be saved to this folder.

A

B

28-9035-38UM Chapter 10, Rev A, 2006

109

Use the Add… function repeatedly to continue adding items to the batch queue until all the analysis runs you want to perform are represented on the list . Checking multiple image stacks in the Image Stack section of the Add Image Stack to Batch Queue window (Figure 10-11A) will submit all of them to the queue when you click OK. In this case, each image stack will be paired with the same analysis protocol (Figure 10-11B).

Each queue entry shows the Image stack file name and the associated analysis protocol file (Figure 10-11B). The status of each entry will be marked as ‘free’ until you run the batch analysis protocol (section 10.5.).

When an item in the queue is highlighted, you can access the following options:

Up arrow—Move selected stack up one position in the queue.

Down arrow—Move selected stack down one position in the queue.

Add—Add an image stack to the analysis queue.

Remove—Remove the selected image stack from the analysis queue.

Reactivate—Reactivate an abandoned analysis.

Done—Exit the dialog box.

Help—Open on-line help

Figure 10-10 An image stack file and associated protocol in the batch queue

28-9035-38UM Chapter 10, Rev A, 2006

110

Figure 10-11 Simultaneous addition of image stacks to a batch queue.(A) Two image stacks are selected for addition to the batch queue; (B) each image stack will be analyzed by the same protocol.

A

B

28-9035-38UM Chapter 10, Rev A, 2006

111

10.5. Running a Batch Analysis

To run a batch analysis, select Application from the main menu and click Image Stack and Analysis… (Figure 10-12). Select Start batch analysis and click OK.The image stack window will appear and analysis will automatically begin on the first image stack in the batch queue (Figure 10-13). As each analysis is complete, the image stack is removed from the batch queue.

Figure 10-12 Starting a batch analysis

Figure 10-13 An image stack running under the Batch Queue Manager

28-9035-38UM Chapter 10, Rev A, 2006

112

At any time during a batch analysis, you can Pause the analysis (Figure 10-14). When the analysis is paused, you can navigate through the image stack and assess the results of analyzed wells as described in Chapter 8.

Figure 10-14 A paused batch analysis showing the activated Batch queue… option

Clicking Batch queue… while the analysis is paused will call up the Batch Queue Manager (Figure 10-15). The status column will indicate the status of each item in the batch queue:

Free—The image stack is waiting to be analyzed.

Being analyzed—Analysis of the image stack is underway.

Abandoned—Analysis has not been completed due to system failure.

Figure 10-15 Batch Queue Manager showing status of image stacks while the batch analysis is paused. The first image stack in the queue is in the process of being analyzed.

28-9035-38UM Chapter 10, Rev A, 2006

113

While the analysis is paused, you can use any of the functions in the Batch Queue Manager window as described in section 10.3. Clicking Done will return you to the Image Stack window.To resume batch analysis, click Analyze in the Image Stack window (Figure 10-14). To abandon the current batch analysis, click Pause to halt the analysis and thenDone to exit the batch analysis. The software will inform you that analysis of the current image stack is incomplete and will offer you the options to keep it in the queue or remove it altogether (Figure 10-16).

Figure 10-16 Warning regarding an unfinished analysis

28-9035-38UM Chapter 10, Rev A, 2006

114

11. Glossary*.cvs, the extension for the analysis results file in a comma- delimited text format .*.dce, the extension for image stack data files from older versions of the IN Cell Analyzer 1000.*.lg2, the extension for the raw analysis output data file.*.run, the extension for image stack data files from older versions of the IN Cell Analyzer 1000.*.txt, the extension for the analysis results file in a tab- delimited text format .*.xann, the extension for the file containing the annotation training set .*.xcls, the extension for the file containing the classification protocol.*.xanp, the extension for the batch queue analysis protocol files.*.xdce, the extension for image stack data files.*.xls, the extension for the analysis results file in MicrosoftExcel format .*.xml, the extension for the analysis results file in an extensible mark-up language format .‘???’, denotes a cell that not been classified. Cells of unknown classification will not be used by the classification protocol builder.1/(Form Factor) (organelle), mean inclusion roundness index (averaged by all inclusions within the cell). Value ranges from 1 to infinity, where 1 is a perfect circle.Analysis mode, mode of operation that allows you to execute analysis using a previously saved analysis protocol. Analysis protocol, a user-defined file that provides the analysis software with information required to extract measurements from images and report measurement results.Analysis, the process of automatically extracting data from specified images using an analysis protocol. This term can also be used to refer to an analysis results file (*.lg2). Annotation, the process of assigning representative cells to defined classes during creation of a classification protocol. Arrow, an Image tool that segments an object based on its local background intensity.Assay development mode, mode of operation that allows you to create, test and edit analysis protocols. Autocontrast, automatic adjustment of a selected image relative to its lightest and darkest pixels, using one of five user-specified mapping functions (Linear, Hyperbola, Logarithmic, Exponential, Spline).Background Intensity (cell), average intensity (in the Cells channel) of pixels in the background immediately adjacent to a cell. The background mask is obtained by dilation of the cytoplasmic outline (not displayed).Batch queue analysis, automated analysis of any number of images stacks using one or more analysis protocols. Batch queue analysis can be particularly useful during analysis protocol optimization, since it allows you to automate repeated analysis of the same image stack using a large number of test protocols. It is also useful for scheduling automated analysis of image stacks from multiple experiments, using a different analysis protocol for each experiment . With batch queue analysis, you can schedule back-to-back overnight analysis runs.Batch Queue Manager, an application that allows you to schedule batch queue analysis.Bitmap, computer-generated outlines that display the results of cell feature identification (segmentation) and classification.

28-9035-38UM Glossary, Rev A, 2006

115

Cell 1/(Form Factor) (cell), mean cell roundness index. Value ranges from 1 to infinity, where 1 is a perfect circle.Cell Area (cell), area of cellCell by cell, mode selected to display analysis results for individual cells.Cell cg X (cell), X coordinate of cell’s center of gravity.Cell cg Y (cell), Y coordinate of cell’s center of gravity.Cell Elongation (cell), mean ratio of the short axis of the cell to the long axis of the cell. If the value is 1 then the object is center-symmetric (not elongated). Short and long axes are orthogonal and not necessarily pass through the shape center of gravity.Cell Gyration Radius (cell), average radius of the shape.Cell Intensity (nucleus), average pixel intensity in the cell body region (within the Nuclei channel).Cell Intensity (cell), average intensity of pixels within the cell body region in the Cells channel.Cell Intensity CV (cell), coefficient of variation of pixel intensities over the population of pixels in the cell body region (within the Cells channel). In Summary mode, the reported value is the average of CVs from all cells in the field of view.Cell/Bckg Intensity (cell), ratio of cell to local background intensity values in the Cells channel.Ch 1, 2, 3, or 4, see Channel.Channel, image data acquired using a specific excitation/emission filter combination. A multi-color image is comprised of data from two or more image channels.Chord ratio (nucleus), shortest chord to longest chord ratio. Both chords pass through the shape’s center of gravity.Circle, an Image tool that allows you to draw a circle or oval with dimensions similar to the object you want to measure, and extract a measurement for the area within the drawn shape.Class, a user-defined label assigned to a phenotypically distinct cell population (e.g. apoptotic, healthy).Classification algorithm, an analysis method used to assign cells to pre-defined classes based on information from a training data set . Also referred to as a classification method. Classification filter, a filter used to classify cells into subpopulations based on user-defined measures. Classification method, see Classification algorithm. Classification protocol builder, an application that allows you to create a classification protocol using data from annotated cells.Classification protocol, an analysis routine (also referred to as a classifier) that uses a specified classification algorithm to assign cells to classes based on information from a training data set . When creating a classification protocol, you specify one of three available classification algorithms. Classification protocols can be saved and later incorporated as sub-routines into any Multi Target Analysis protocol. Classifier, see Classification protocol.Collar, a segmentation method that establishes a cytoplasmic sampling region by dilating outwards a defined distance from the established nuclear region. Use this method if you need to sample cytoplasmic intensity rapidly (for example, in a kinetic assay).Color Bar Properties, see Color BarColor Bar, an application that allows you to color-code wells of the plate map to indicate the results of analysis. Color- coding can be set to reflect the output measure of choice (e.g. nuclear area, inclusion/cell intensity).

28-9035-38UM Glossary, Rev A, 2006

116

Compactness (nucleus), compactness is another measure characterizing shape. It is calculated by the formula: 2*PI*(gyration radius*gyration radius2)/area; Gyration radius is an average radius of a shape.Count (organelle), number of inclusions attributed to a cell (in Cell by Cell mode), or averaged for all cells in the image (in Summary mode).CVA scatterplot, a graph within the Classification Protocol Builder that helps you visualize how effectively defined cell classes comprising the training data set are resolved in multi-dimensional feature space using the selected features. Canonical variance analysis (CVA) is used to analyze multi- dimensional data. The results are then collapsed into two dimensions, and graphed on a scatter-plot .Decision tree, a classification tool consisting of multiple Threshold or Linear Discriminant 2D scatter plot filters to classify cells into subpopulations. Dimensionality, see Sub-space dimensionality.Distance to Nuc (organelle), mean distance from the inclusion center of gravity to the nucleus center of gravity (averaged by all inclusions within the cell).Elongation (organelle), mean ratio of the short axis of the inclusion to the long axis of the inclusion (averaged by all inclusions within the cell). If the value is 1 then the object is center-symmetric (not elongated).Exponential, an Autocontrast mapping function: intensity values are mapped using an exponential scale, giving greater differentiation at higher pixel intensities.Feature Space, a multi-dimensional space where each cell is represented as a single point . Each specified measurement (‘feature’) from the cell gives the coordinate of the point along one axis of the space. The dimensionality of the feature space is thus equal to the number of features used. For example, if two features are used, the space will be a plane, with the first feature on the X-axis, and the second feature on the Y-axis.Filter, a set of one or more thresholds applied after primary analysis to identify objects that have measures above or below user-specified values, or that fall within a specified range. Optionally, a filter can be used to eliminate cells from analysis.Global Threshold, a segmentation method for nuclei that provides a means of identifying objects based on intensity values and is independent of object size and shape. Goodness (KS pValue), measure of how well a given set of data fits the non-parametric Kolmogorov-Smirnov experimental model. The larger the value, the better the data fit the model.Goodness (Metric), metric measure of inter-class separation in multi-dimensional feature space, based on inter-class to intra-class distance ratio. The larger the value, the better the class separation.Granule, see inclusion.Gray level, dimensionless unit denoting the relative brightness of a pixel in a monochrome image. IN Cell Analyzer 1000 monochrome images have a gray level range from 0 to 4096, where a pixel with a value of 0 is completely black and a pixel with a value of 4096 is completely white.See also Relative fluorescent units.Hyperbola, an Autocontrast mapping function: intensity values are mapped using a hyperbolic function. This is particularly useful to view dim images, since the curve can be distorted to enhance the lower end of the intensity range. Image acquisition, the process of capturing images from samples (usually on multi-well plates) using IN Cell Analyzer 1000. The output file of raw results is denoted by the extension *.xdce.Image Stack, the set of images obtained from one plate by image acquisition on IN Cell Analyzer 1000.Inclusion, groups of contiguous pixels within a defined cell region that are a specified degree brighter than the image background (e.g. mitochondria, condensed DNA, vesicles). Inclusions are sometimes also referred to as granules.

28-9035-38UM Glossary, Rev A, 2006

117

Inclusion/Bckg Intensity (organelle), inclusion to background intensity ratio. Background intensity is measured immediately next to the cell. (Both Inclusion and Background values are obtained in the Organelles channel). Inclusion/Cell Intensity (organelle), ratio of the average pixel intensity within inclusions to the average pixel intensity in the cell region immediately adjacent to the inclusions (both intensities measured in the Organelles channel). Intensity (N+C) (nucleus) average intensity in whole cell (nucleus + cytoplasm) (Nuclei channel).Intensity (N+C) (cell), average intensity in whole cell (nucleus + cytoplasm) (Cells channel).Intensity (organelle), average intensity of pixels within inclusions.Intensity Spreading (cell), indicator of intensity distribution within the object (cell).If the value is 1, then the intensity is distributed uniformly across the object (cell). If the value is > 1, the intensity concentrates near the periphery of the object . If the value is < 1, the intensity concentrates near the center of the object . Integrated Intensity (Nuc) (nucleus), the amount of light emitted by the nucleus. It is equal to nucleus average intensity (Nuclei channel) multiplied by nucleus area. Column title in Summary tables of results is IxA (Nuc).Integrated Intensity (Cell) (nucleus), the amount of light emitted by the cytoplasm. It is equal to cytoplasm average intensity (Nuclei channel) multiplied by cytoplasm area. Column title in Summary tables of results is IxA (Cell)Integrated Intensity (whole cell) (nucleus), the amount of light emitted by the whole cell. It is equal to cell average intensity (Nuclei channel) multiplied by cell area. Column title in Summary tables of results is IxA (N+C)Integrated Intensity (Nuc) (cell), the amount of light emitted by the nucleus. It is equal to nucleus average intensity (Cells channel) multiplied by nucleus area. Column title in Summary tables of results is IxA (Nuc).Integrated Intensity (Cell) (cell), the amount of light emitted by the cytoplasm. It is equal to cytoplasm average intensity (Cells channel) multiplied by cytoplasm area. Column title in Summary tables of results is IxA (Cell)Integrated Intensity (whole cell) (cell), the amount of light emitted by the whole cell. It is equal to cell average intensity (Cells channel) multiplied by cell area. Column title in Summary tables of results is IxA (N+C)Length: linear, an Image tool that measures the straight-line distance between two points.Length: trace, an Image tool that is unavailable with the Multi Target Analysis analysis module.Light Flux (relative) (nucleus), the normalized amount of light emitted by the whole nuclei. It is equal to nucleus average intensity multiplied by area and normalised by cytoplasm average intensity (Nuclei channel). Nuclei area is taken in pixels. Column title in Summary tables of results is Light Flux.Light Flux (relative) (cell), the normalized amount of light emitted by the whole nuclei. It is equal to nucleus average intensity multiplied by area and normalized by cytoplasm average intensity (Cells channel ). Nucleus area is taken in pixels. Column title in Summary tables of results is Light Flux.Linear, an Autocontrast mapping function: intensity values are mapped on a linear scale. The signal input (original pixel intensity) is directly proportional to the signal output (displayed pixel intensity).Linear Discriminant 2D filter, a 2D scatter plot filter used to classifiy cells into multiple subpopulations based on two different measures. Can be used independently or within a Decision tree. Logarithmic, an Autocontrast mapping function: intensity values are mapped using an logarithmic scale, giving greater differentiation at lower pixel intensities.Look-up table, see LUT.

28-9035-38UM Glossary, Rev A, 2006

118

LUT, a look-up table (LUT) is a table of cross-references linking index numbers to output values. The LUT is used to determine the intensity values and/or colors with which a particular image will be displayed. Three pre-defined LUTs are loaded when IN Cell Analyzer 1000 Workstation starts: MONOCHRM.VIS, a monotonically increasing gray scale LUT; REVERSE.VIS, a monotonically decreasing gray scale LUT; and SPECT2.VIS, a pseudo-color LUT which allows easier discrimination of relatively small changes in intensity. More LUTs may be added to this list box with the File – Open and File – Import menu items.Magnification changer, a control on IN Cell Analyzer 1000 that allows magnification to be amplified 1.5 times when set to the 1.5X position. The default position after installation is 1X (i.e. no extra magnification). The analysis workstation ruler tool will be accurate only if the magnification changer is set to 1X during image acquisition.Mean Area (organelle), mean area of inclusions (averaged by all inclusions within the cell).Measure, data which can be obtained from each image (e.g. cell count , nuclei intensity).MONOCHROM.VIS, see LUT.Multiscale top-hat, a segmentation method that provides a means of identifying objects that come in a range of different sizes. Choose this method if the objects you want to identify are heterogeneous in size and shape.Nearest neighbors, a cell classification algorithm considers the five nearest classified points in multi-dimensional feature space, and assigns the cell in question to the majority class. Try this method if the classes have complex phenotypes that are likely to be difficult to distinguish.Neighbor count (SOI) (nucleus), number of neighbouring nuclei, determined using the Sphere Of Influence (SOI) adjacency rule. Neighbor count (MIN) (nucleus), number of neighbouring nuclei, determined using the MIN rule.Neighbor count (Gabriel) (nucleus), number of neighbouring nuclei, determined using the Gabriel rule.Neighbor count (Lune) (nucleus), number of neighbouring nuclei, determined using the Lune rule.Neighbor Count (organelle), number of neighboring inclusions, determined using the Sphere Of Influence adjacency rule (averaged by all inclusions within the cell). Neural network, a classification algorithm that examines the training data set to ‘learn’ how to handle complex data. For the learning element of the algorithm to work well, large numbers of cells must be annotated for each class. Try this method if the classes are likely to be difficult to distinguish, but you have annotated a large number of cells for each class.Noise correction, reduces image noise (unpredictable or random signal) prior to further image processing steps. Try noise correction if the cytoplasmic signal is weak relative to the surrounding background intensity.Noise, see Noise correction.Nuclei coordination, refers to the spatial coordinate measurements based on adjacency rules for the determination of whether objects are considered neighbors or not.Nuc Area (nucleus), area of identified nucleus.Nuc/Cell Area (cell), ratio of nucleus to cell area (Cells channel).Nuc cg X (nucleus), X coordinate of nucleus’s center of gravity.Nuc cg Y (nucleus), Y coordinate of nucleus’s center of gravity.Nuc Displacement (nucleus), nuclear displacement is the distance between the nucleus’s and the cell’s centers of gravity, normalized by the gyration radius of the nucleus.

28-9035-38UM Glossary, Rev A, 2006

119

Nuc Elongation (nucleus), mean ratio of the short axis of the nucleus to the long axis of the nucleus. If the value is 1 then the object is center- symmetric (not elongated).Nuc 1/(Form Factor), mean nucleus roundness index. Value ranges from 1 to infinity, where 1 is a perfect circle.Nuc Intensity (nucleus), average nuclear intensity.Nuc Intensity (cell), average intensity of pixels within the nuclear region (in the Cells channel).Nuc Intensity CV (nucleus), coefficient of variation of pixel intensities over the population of pixels comprising the nuclear region. In Summary mode, the reported value is the average of the CVs from all nuclei in the field of view. Nuc Intensity CV (cell), coefficient of variation of pixel intensities over the population of pixels in the nuclear region within the Cells channel. In Summary mode, the reported value is the average of CVs from all nuclei in the field of view. Nuc/Cell Intensity (nucleus), ratio of intensities sampled in the nuclear and cell body regions.Nuc/Cell Intensity (cell), ratio of nuclear to cytoplasmic intensity values (both values obtained from the Cells channel).Optimal Sub-space, the multi-dimensional feature space identified as optimal for discrimination of cell classes comprising the training data set when using the Find Optimal Sub-Space function of the Classification Protocol Builder.Outline, an Image tool that allows you to draw a freehand outline around the representative feature of interest . Overlays, see Bitmap.Protocol, see Analysis protocol.Pseudo, a segmentation method that applies a copy of a specified, previously-defined object mask (Nuclei, Cells or Organelles) to an additional Reference channel. Quadratic discriminant, a classification algorithm that uses a parametric probabilistic classification model, producing non-linear (second-order) curved boundaries between classes in feature space. This method is appropriate when each defined cell class is homogeneous in appearance.Reference, an optional object type that can be selected to obtain intensity measurements from a specified region from the source image of choice. Region growing, a segmentation method that establishes the object boundaries by incrementally dilating the defined region until the background (relatively dim pixels) is encountered. Use this method if the definition of the object boundaries is critical for your assay. Relative fluorescent unit (RFU). Dimensionless unit corresponding to the relative intensity of a fluorescent signal. The IN Cell Analyzer 1000 camera capturesfluorescent signal as a monochrome image, which has gray levels (RFU) ranging from 0 to 4096.REVERSE.VIS, see LUT.RFU, see Relative fluorescent units (RFU).Scale, size scale at which an image is evaluated during the process of object segmentation. The number of scales you specify for the multi-scale top-hat segmentation method determines the number of size scales (within the specified size range) at which the software will apply top hat transformation.Segmentation, the process of dividing an image into a number of individual objects or contiguous regions, differentiating them from each other and from the image background.

28-9035-38UM Glossary, Rev A, 2006

120

Sensitivity, a threshold applied during feature segmentation that determines which pixel clusters will qualify as objects (nuclei, inclusions, or cells) based on their intensity relative to local background. The lower the sensitivity setting, the brighter a cluster of pixels must be in order to be differentiated from the background; detection of dimmer objects will increase as the percent sensitivity increases. Shading removal, reduces unwanted shading (uneven intensity) in an image prior to further image processing steps. Shading may be caused by non-uniform sample illumination, non-uniform camera sensitivity, or dirt/dust on lens surfaces.Shading, see Shading removal.Spacing (SOI) (nucleus), measure of the inter-nuclear distance (distance from the center of gravity to the neighbor’s center of gravity, averaged by all neighboring nuclei), determined using the Sphere Of Influence (SOI) adjacency rule. Spacing (MIN) (nucleus), measure of the inter-nuclear distance, determined using the MIN rule.Spacing (Gabriel) (nucleus), measure of the inter-nuclear distance, determined using the Gabriel rule.Spacing (Lune) (nucleus), measure of the inter-nuclear distance, determined using the Lune rule.Spacing (organelle), measure of the inter-inclusion distance, determined using the Sphere Of Influence (SOI) adjacency rule (averaged by all inclusions within the cell).SPECT2.VIS, see LUT.Spline, an Autocontrast mapping function: intensity values are mapped using a cubic spline scale. The cubic spline is a flexible way to manipulate segments of the intensity spectrum.Stack, see Image Stack.Sub-space dimensionality, the number of features specified for class discrimination when using the Define Optimal Sub- Space function within the Classification Protocol Builder.Sub-space, see Optimal Sub-space.Summary, mode selected to report analysis measurements averaged for the population of cells identified within an image or well.Supervised classification, see Classification.Threshold filter, a 1D filter that allows cells to be filtered or classified based on a single measure. Can be used independently or within a Decision tree. Time Lapse Stack, an application that allows you to specify what the software will do if images are missing from a time series.Top-hat, a rapid segmentation method that is used to accentuate objects of a specified size. Top-hat transformation is useful for distinguishing objects from a surrounding uneven background.Total Area (organelle), total area of inclusions attributed to the cell.Training data set, a file containing cell annotation data. The training data set is used to create a classifier that will automatically assign cells to defined classes.Values, an Annotation Tool window page that displays analysis measurements for the cell being annotated. Use this information to help you decide to which class a cell should be assigned.Wave, each wave channel (Wave 1 to Wave 4) corresponds to a different emission wavelength.

28-9035-38UM Glossary, Rev A, 2006

imagination at work

http://www.gehealthcare.com/lifesciences GE Healthcare UK Limited Amersham Place, Little Chalfont,Buckinghamshire, HP7 9NA UK

28-9035-38UM, Rev A, 2006

GE Healthcare offi ces:

GE Healthcare Bio-Sciences ABBjörkgatan 30 751 84 Uppsala Sweden

GE Healthcare Europe GmbH Munzinger Strasse 5 D-79111 Freiburg Germany

GE Healthcare UK LimitedAmersham Place Little Chalfont BuckinghamshireHP7 9NA UK

GE Healthcare Bio-Sciences Corp800 Centennial Avenue P.O. Box 1327 PiscatawayNJ 08855-1327 USA

GE Healthcare Bio-Sciences KK Sanken Bldg. 3-25-1Hyakunincho Shinjuku-kuTokyo 169-0073 Japan

GE Healthcare regional offi ce contact numbers:

Asia Pacifi cTel: +85 65 6 275 1830Fax: +85 65 6 275 1829

AustralasiaTel: + 61 2 8820 8299Fax: +61 2 8820 8200

AustriaTel: 01/57606-1613Fax: 01/57606-1614

BelgiumTel: 0800 73 890Fax: 02 416 82 06

CanadaTel: 1 800 463 5800Fax: 1 800 567 1008

Central, East, & South East EuropeTel: +43 1 972 720Fax: +43 1 972 722 750

DenmarkTel: 45 70 25 24 50Fax: 45 45 16 2424

EireTel: 0 1800 709992Fax: +44 1494 542010

Finland & BalticsTel: +358 9 512 39 40Fax: +358 9 512 39 439

FranceTel: 01 6935 6700Fax: 01 6941 9877

GermanyTel: 0800 9080 711Fax: 0800 9080 712

Greater ChinaTel: +852 2100 6300Fax: +852 2100 6338

ItalyTel: 02 26001 320Fax: 02 26001 399

JapanTel: 81 3 5331 9336Fax: 81 3 5331 9370

KoreaTel: +82 2 6201 3700Fax: +82 2 6201 3803

Latin AmericaTel: +55 11 3933 7300Fax: + 55 11 3933 7304

Middle East & AfricaTel: +30 210 9600 687Fax: +30 210 9600 693

NetherlandsTel: 0800 8282821Fax: 0800 8282824

NorwayTel: 815 65 777Fax: 815 65 666

PortugalTel: 21 417 7035Fax: 21 417 3184

Russia, CIS & NIS Tel: +7 495 956 5177Fax: +7 495 956 5176

SpainTel: 902 11 72 65Fax: 935 94 49 65

SwedenTel: 018 612 1900Fax: 018 612 1910

SwitzerlandTel: 0848 8028 12Fax: 0848 8028 13

UKTel: 0800 515 313Fax: 0800 616 927

USATel: +1 800 526 3593Fax: +1 877 295 8102