written by: william zimmerman aaron logan dylan reid william lim kyungchul song i.r. s n a pp (image...

13
Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application ) May09-10

Upload: moses-goodwin

Post on 11-Jan-2016

220 views

Category:

Documents


1 download

TRANSCRIPT

Page 1: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

Written By:

William Zimmerman

Aaron Logan

Dylan Reid

William Lim

Kyungchul Song

I.R. S N A pp(Image Reconstruction and Segmentation for Neurosurgery Application )

May09-10

Page 2: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

Problem / need statement

Develop an algorithm for Compressed Sensing  Compressed Sensing for MRI  Batch Compress Sensing for Dynamic MRI

 Develop an algorithm for Sequential Segmentation(Currently, MRI data is segmented either by hand, or a very slow algorithm.)

To be able to sequentially segment deforming objects or Regions of Interest (ROI's) from the filtered, compressed images.

Utilize prior knowledge about shape change dynamics to segment noisy/low

contrast imagery.

Make this process fast enough to run in real-time, using only current and past images for segmentation.

Page 3: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

Concept sketch / mockup

Example GUI’S

Processing GUI

Segmentation GUI

Page 4: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

System Description(1/3)

Page 5: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

System Description(2/3)

Page 6: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

System Description(3/3)

Page 7: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

Operating Environment

In summary, these are the attributes of the operating environment:

Linux-based

Fast, multiple-core processing

Lots of memory available

Programs interfaced through command prompt program

Page 8: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

User Interface Description

The preliminary user interface will consist of running the C++

program in a command prompt window. After the code is closer

to a working product, we will create an executable that has a

graphical user interface (GUI) that will allow the user to pick

different options related to segmentation, and perhaps allow the

user to assist the process if necessary. See figure 1 and 2 for

general mock-ups

Page 9: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

Non / Functional Requirement

Functional Requirements 

FR001: To translate a matlab algorithm written by Dr. Vaswani’s graduate students, to C++.

FR002: To run experiments with actual MRI data. These experiments will include testing of

the  compression algorithm and the sequential segmentation algorithm.

FR003: To check the correctness of our output. This will be tested by comparing output data

from the matlab code to the output of our C++ translation.

 

Non-functional Requirements: 

NFR001: The program shall be written in C++.

NFR002: The program shall run faster than in matlab.

NFR003: The program shall be capable of running in real time.

Page 10: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

Deliverables

The deliverables of our project include

C++ code, well documented and working

An executable of the C++ code

Test data and results from all tests

A complete system that takes raw MR data, first reconstructs the image, and

then performs segmentation and outputs the contour

A basic GUI that may allow the user to assist the segmentation process, and

also displays segmentation options

Page 11: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

Market / literature survey

Magnetic resonance imaging (MRI) is an imaging technique used

in medical that has a strong market in medical imaging industry.

Area such as Radiology uses MRI to visualize structure of body.

Images from MRI can be utilized for guiding invasive surgery.

This is due to the ability to image soft tissue and organs.

Page 12: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

Deliverables

The deliverables of our project include:

C++ code, well documented and working

An executable of the C++ code

Test data and results from all tests

A complete system that takes raw MR data, first reconstructs the image, and

then performs segmentation and outputs the contour

A basic GUI that may allow the user to assist the segmentation process, and

also displays segmentation options

Page 13: Written By: William Zimmerman Aaron Logan Dylan Reid William Lim Kyungchul Song I.R. S N A pp (Image Reconstruction and Segmentation for Neurosurgery Application

R i s k s

Risks with this project mainly relate to the segmentation part.

For one, segmentation is very data dependent, and that the data

may be too general to output an accurate contour. Also, another

risk is that running the KFCS algorithm on a basic PC could

cause it to crash, because of the high amount of memory needed

to run it.