integrating deep learning platforms within enterprise ... · barbaros s. erdal, ph.d. department of...

31
Confidential │ Trade Secret │ Proprietary │ Do Not Copy 1 Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 2017 Integrating Deep Learning Platforms within Enterprise Level Medical Imaging Environments Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017

Upload: others

Post on 19-Aug-2020

1 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

1

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 2017

Integrating Deep Learning Platforms within Enterprise Level

Medical Imaging Environments

Barbaros S. Erdal, Ph.D.

Department of Radiology

The Ohio State University Wexner Medical Center

May, 2017

Page 2: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

2

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 20172

REMIX Developmentin support of

Images

ImagesImages

TCC

Clinical

Tissue

Molecular

Molecular

Molecular

Clinical

Clinical

Tissue

Tissue

Page 3: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

3

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 20173

Quantitative Imaging is Essential in Cancer Care:

Tumor Characteristics:

• Size (+):

o 2-D: RECIST 1.1

o 3-D: Volumetric Assessment

o Potential: Shape / Peri-Tumor Zone / Relationships - ?3D Printing?

• Histopathology:

o Tissue Heterogeneity: e.g. Textural Analysis

o Tissue Mechanical Properties: e.g. Stiffness by MR

• Pathophysiology:

o Metabolic Abnormalities: MR Spectroscopy / PET

o Perfusion Abnormalities: MRI / SPECT / PET / CT

• Future:

o New Image-Data Reconstructions

Robust Conventional Imaging Data: e.g. MR “Fingerprinting”

Raw Imaging Data: Collaboration with Imaging Industry

o New Imaging Acquisitions with Established Technologies: e.g. CT

Page 4: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

4

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 20174

Tissue Specimens

Molecular Data

Longitudinal Clinical Data

For each Patient collect:

Quantitative Imaging

Qualitative

Imaging

Total Cancer Care Protocol: An Opportunity for OSU

Page 5: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

5

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 20175

CT Gray Level Texture Analysis as a Quantitative Imaging Biomarker for Epidermal

Growth Factor Receptor Mutation Status in Adenocarcinoma of the Lung. [Ozkan E, et al. Am J Roentgenol 2015]

Radiomics

Page 6: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

6

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 20176

REMIX: RIS/PACS Workflow

EMR System

AgeGenderReason

CPOE for Radiology Exam

Physician requests imagingstudy

Imaging Exam Requesting

Facility

REMIX

HIS/RIS

Exam Scheduled and Performed

PACSReconstructed images are searchable, and ready foradvanced Image analysis

All study relevant data and Images are linkable andmineable

Clinical Reasearchers

REMIX receives and de-identifiesimage data and related metadata

Data Warehouse

Patients have already beenConsented for TCCP

Page 7: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

7

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 20177

Physician requests imagingstudy

Imaging Exam Requesting

Facility

REMIX Desktop

PACS

Reconstructed images are searchable, and ready foradvanced Image analysis

All study relevant data, Images are linkable andmineable

Clinical Reasearchers / Researchers

REMIX receives and de-identifiesimage data and related metadata

Scanner

Enterprise Data Warehouse

Patients have already beenConsented based on study involved

Diagnostic Workstation

EMR\HIS\RISEnterprise Viewer

VNA

REMIX Recon REMIX BIREMIX AI

REMIX Server

Clinical\Operational User

Clinical and Operationaluses are uninterrupted

• REMIX

Page 8: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

8

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 20178

• REMIX: De-Identification (Operated by OSUWMC Imaging Informatics)

Patient Search for Research Dataset Preparation

Clinical Trial Support

Honest Broker Compliant Batch Image Processing for Large datasets

Data verification supported by OSUWMC Imaging Informatics

CD Burning and Image sending to custom folders and\or destinations

Page 9: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

9

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 2017

9Future Directions: Data Access, Radiology / Imaging

Source: OSUP Finance (Epic Cadence – Internal)

Page 10: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

10

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201710

REMIX Data-Mining Capabilities

Page 11: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

11

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201711

REMIX Data-Mining Capabilities

Page 12: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

12

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201712

ORIEN OSU PATIENT 2:

• 63YO Male

• Smoking History

• Sudden Weight Loss

ROQID

REMIX Interactive Capabilities

Page 13: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

13

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201713

REMIX Pathology

Page 14: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

14

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201714

REMIX: Quantitative Capabilities

Page 15: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

15

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201715

2-D Quantitative Capabilities

Page 16: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

16

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201716

3-D Quantitative Capabilities

Page 17: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

17

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201717

Developed Texture-Analysis Capabilities

Page 18: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

18

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201718

Texture Analysis Capabilities

Page 19: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

19

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201719

Explore “Deep Learning” for Pattern Recognition inImages, Digital Pathology, and Genomic Data:

Page 20: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

20

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 2017

20What is New in the Field: Radiology

Positive

(Gold standard)

Negative

(Gold Standard)

Total

Positive

(AI algorithm)

38 5 43

Negative

(AI Algorithm)

2 35 37

Total 40 40 80

Neurologic Disease Medical Imaging Informatics

(e.g., Artificial Intelligence)

Cardiovascular DiseaseFast MRI

(e.g., MRE, 4D Flow)

Cancer Low-Dose MolecuIar Imaging

(e.g., Digital PET)

Standard 10 x Reduction

NCI R01CA195513

RSNA Medical Student

RSNA Molecular Imaging

Ohio Third Frontier

Page 21: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

21

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201721

Physician requests imagingstudy

Imaging Exam Requesting

Facility

REMIX Desktop

PACS

Reconstructed images are searchable, and ready foradvanced Image analysis

All study relevant data, Images are linkable andmineable

Clinical Reasearchers / Researchers

REMIX receives and de-identifiesimage data and related metadata

Scanner

Enterprise Data Warehouse

Patients have already beenConsented based on study involved

Diagnostic Workstation

EMR\HIS\RISEnterprise Viewer

VNA

REMIX Recon REMIX BIREMIX AI

REMIX Server

Clinical\Operational User

Clinical and Operationaluses are uninterrupted

• REMIX

Page 22: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

22

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 2017

22Quality Metrics: Radiology - Clinical Service Efficiency

Source: IHIS

Neuro MRI: Routine vs Stat

Stat

Role for Artificial Intelligence in Imaging

Page 23: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

23

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201723

Positive

(Gold standard)

Negative

(Gold Standard)

Total

Positive

(AI algorithm)

38 5 43

Negative

(AI Algorithm)

2 35 37

Total 40 40 80

Neurologic Disease Medical Imaging Informatics

(e.g., Artificial Intelligence)

REMIX -AI

Page 24: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

24

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201724

REMIX - Desktop

Page 25: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

25

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201725

REMIX - Recon

Page 26: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

26

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201726

Pitfalls

Page 27: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

27

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201727

REMIX AI specs

Processor 1x intel Core i7-5930 K Processor (15M Cache,

3.50 GHz)

Memory 64GB DDR4

GPUs 4 x NVIDIA GeForce GTX Titan X GPUs (7

Teraflops of single precision, 336.5 GB/s of

memory bandwidth, 12 GB memory per GPU)

Operating System (OS) Ubuntu 14.04

Storage 2x 256 GB SSD disk foe

r OS and software libraries and 3x3TB standard

disk on RAID 5 for data storage

Connecting to REMIX AI:1) REMIX AI web interface, allowing users to upload their data into the

system2) REMIX Desktop, permitting users to directly save their image data

into shared disk drives of REMIX AI3) Python-based client libraries, so that users can make Restful API calls

to REMIX PACS

Page 28: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

28

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201728

Example

Page 29: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

29

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201729

REMIX AI Performance

Images to 256x256 matrix and processed with GoogLeNet convolutional network running on Caffe. 60 training epochs used.

Model creation with the first dataset (from Query 1, with 2,583 images) was 6 minutes and 19 seconds

Model creation for the second dataset (from Query 2, with 646 images), total processing took 97 seconds

Once image-classification models were created, batch image classifications performed at approximately 25 images per second.

Page 30: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

30

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201730

Physician requests imagingstudy

Imaging Exam Requesting

Facility

REMIX Desktop

PACS

Reconstructed images are searchable, and ready foradvanced Image analysis

All study relevant data, Images are linkable andmineable

Clinical Reasearchers / Researchers

REMIX receives and de-identifiesimage data and related metadata

Scanner

Enterprise Data Warehouse

Patients have already beenConsented based on study involved

Diagnostic Workstation

EMR\HIS\RISEnterprise Viewer

VNA

REMIX Recon REMIX BIREMIX AI

REMIX Server

Clinical\Operational User

Clinical and Operationaluses are uninterrupted

• REMIX

Page 31: Integrating Deep Learning Platforms within Enterprise ... · Barbaros S. Erdal, Ph.D. Department of Radiology The Ohio State University Wexner Medical Center May, 2017. Confidential

Confidential │ Trade Secret │ Proprietary │ Do Not Copy

31

Confidential │ Trade Secret │ Proprietary │ Do Not Copy Strategy and Planning │The Ohio State University Wexner Medical Center © 201731

Questions?

Barbaros S. Erdal, Ph.D.

Department of RadiologyThe Ohio State University Wexner Medical Center

Contact: [email protected]