Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging

Varun A. Kelkar; Dimitrios S. Gotsis; Frank J. Brooks; Kyle J. Myers; K. C. Prabhat; Rongping Zeng; Mark A. Anastasio

doi:10.1117/12.2612893

Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging

Varun A. Kelkar, Dimitrios S. Gotsis, Frank J. Brooks, Kyle J. Myers, K. C. Prabhat, Rongping Zeng, Mark A. Anastasio

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Abstract

Modern generative models, such as generative adversarial networks (GANs), hold tremendous promise for several areas of medical imaging, such as unconditional medical image synthesis, image restoration, reconstruction and translation, and optimization of imaging systems. However, procedures for establishing stochastic image models (SIMs) using GANs remain generic and do not address specific issues relevant to medical imaging. In this work, canonical SIMs that simulate realistic vessels in angiography images are employed to evaluate procedures for establishing SIMs using GANs. The GAN-based SIM is compared to the canonical SIM based on its ability to reproduce those statistics that are meaningful to the particular medically realistic SIM considered. It is shown that evaluating GANs using classical metrics and medically relevant metrics may lead to different conclusions about the fidelity of the trained GANs. This work highlights the need for the development of objective metrics for evaluating GANs.

Original language	English (US)
Title of host publication	Medical Imaging 2022
Subtitle of host publication	Image Perception, Observer Performance, and Technology Assessment
Editors	Claudia R. Mello-Thoms, Claudia R. Mello-Thoms, Sian Taylor-Phillips
Publisher	SPIE
ISBN (Electronic)	9781510649453
DOIs	https://doi.org/10.1117/12.2612893
State	Published - 2022
Externally published	Yes
Event	Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment - Virtual, Online Duration: Mar 21 2022 → Mar 27 2022

Publication series

Name	Progress in Biomedical Optics and Imaging - Proceedings of SPIE
Volume	12035
ISSN (Print)	1605-7422

Conference

Conference	Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment
City	Virtual, Online
Period	3/21/22 → 3/27/22

Keywords

Generative adversarial networks
image perception
objective image quality assessment
stochastic image models

ASJC Scopus subject areas

Electronic, Optical and Magnetic Materials
Atomic and Molecular Physics, and Optics
Biomaterials
Radiology Nuclear Medicine and imaging

Online availability

10.1117/12.2612893

Library availability

Discover UIUC Full Text

Cite this

Kelkar, V. A., Gotsis, D. S., Brooks, F. J., Myers, K. J., Prabhat, K. C., Zeng, R., & Anastasio, M. A. (2022). Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging. In C. R. Mello-Thoms, C. R. Mello-Thoms, & S. Taylor-Phillips (Eds.), Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment Article 120350O (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 12035). SPIE. https://doi.org/10.1117/12.2612893

Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging. / Kelkar, Varun A.; Gotsis, Dimitrios S.; Brooks, Frank J. et al.
Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment. ed. / Claudia R. Mello-Thoms; Claudia R. Mello-Thoms; Sian Taylor-Phillips. SPIE, 2022. 120350O (Progress in Biomedical Optics and Imaging - Proceedings of SPIE; Vol. 12035).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution

Kelkar, VA, Gotsis, DS, Brooks, FJ, Myers, KJ, Prabhat, KC, Zeng, R & Anastasio, MA 2022, Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging. in CR Mello-Thoms, CR Mello-Thoms & S Taylor-Phillips (eds), Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment., 120350O, Progress in Biomedical Optics and Imaging - Proceedings of SPIE, vol. 12035, SPIE, Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment, Virtual, Online, 3/21/22. https://doi.org/10.1117/12.2612893

Kelkar VA, Gotsis DS, Brooks FJ, Myers KJ, Prabhat KC, Zeng R et al. Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging. In Mello-Thoms CR, Mello-Thoms CR, Taylor-Phillips S, editors, Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment. SPIE. 2022. 120350O. (Progress in Biomedical Optics and Imaging - Proceedings of SPIE). doi: 10.1117/12.2612893

Kelkar, Varun A. ; Gotsis, Dimitrios S. ; Brooks, Frank J. et al. / Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging. Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment. editor / Claudia R. Mello-Thoms ; Claudia R. Mello-Thoms ; Sian Taylor-Phillips. SPIE, 2022. (Progress in Biomedical Optics and Imaging - Proceedings of SPIE).

@inproceedings{40a6a7d23e60404d9872ee26867c28cf,

title = "Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging",

abstract = "Modern generative models, such as generative adversarial networks (GANs), hold tremendous promise for several areas of medical imaging, such as unconditional medical image synthesis, image restoration, reconstruction and translation, and optimization of imaging systems. However, procedures for establishing stochastic image models (SIMs) using GANs remain generic and do not address specific issues relevant to medical imaging. In this work, canonical SIMs that simulate realistic vessels in angiography images are employed to evaluate procedures for establishing SIMs using GANs. The GAN-based SIM is compared to the canonical SIM based on its ability to reproduce those statistics that are meaningful to the particular medically realistic SIM considered. It is shown that evaluating GANs using classical metrics and medically relevant metrics may lead to different conclusions about the fidelity of the trained GANs. This work highlights the need for the development of objective metrics for evaluating GANs.",

keywords = "Generative adversarial networks, image perception, objective image quality assessment, stochastic image models",

author = "Kelkar, {Varun A.} and Gotsis, {Dimitrios S.} and Brooks, {Frank J.} and Myers, {Kyle J.} and Prabhat, {K. C.} and Rongping Zeng and Anastasio, {Mark A.}",

note = "Publisher Copyright: {\textcopyright} 2022 SPIE. All rights reserved.; Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment ; Conference date: 21-03-2022 Through 27-03-2022",

year = "2022",

doi = "10.1117/12.2612893",

language = "English (US)",

series = "Progress in Biomedical Optics and Imaging - Proceedings of SPIE",

publisher = "SPIE",

editor = "Mello-Thoms, {Claudia R.} and Mello-Thoms, {Claudia R.} and Sian Taylor-Phillips",

booktitle = "Medical Imaging 2022",

}

TY - GEN

T1 - Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging

AU - Kelkar, Varun A.

AU - Gotsis, Dimitrios S.

AU - Brooks, Frank J.

AU - Myers, Kyle J.

AU - Prabhat, K. C.

AU - Zeng, Rongping

AU - Anastasio, Mark A.

PY - 2022

Y1 - 2022

N2 - Modern generative models, such as generative adversarial networks (GANs), hold tremendous promise for several areas of medical imaging, such as unconditional medical image synthesis, image restoration, reconstruction and translation, and optimization of imaging systems. However, procedures for establishing stochastic image models (SIMs) using GANs remain generic and do not address specific issues relevant to medical imaging. In this work, canonical SIMs that simulate realistic vessels in angiography images are employed to evaluate procedures for establishing SIMs using GANs. The GAN-based SIM is compared to the canonical SIM based on its ability to reproduce those statistics that are meaningful to the particular medically realistic SIM considered. It is shown that evaluating GANs using classical metrics and medically relevant metrics may lead to different conclusions about the fidelity of the trained GANs. This work highlights the need for the development of objective metrics for evaluating GANs.

AB - Modern generative models, such as generative adversarial networks (GANs), hold tremendous promise for several areas of medical imaging, such as unconditional medical image synthesis, image restoration, reconstruction and translation, and optimization of imaging systems. However, procedures for establishing stochastic image models (SIMs) using GANs remain generic and do not address specific issues relevant to medical imaging. In this work, canonical SIMs that simulate realistic vessels in angiography images are employed to evaluate procedures for establishing SIMs using GANs. The GAN-based SIM is compared to the canonical SIM based on its ability to reproduce those statistics that are meaningful to the particular medically realistic SIM considered. It is shown that evaluating GANs using classical metrics and medically relevant metrics may lead to different conclusions about the fidelity of the trained GANs. This work highlights the need for the development of objective metrics for evaluating GANs.

KW - Generative adversarial networks

KW - image perception

KW - objective image quality assessment

KW - stochastic image models

UR - http://www.scopus.com/inward/record.url?scp=85130012056&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85130012056&partnerID=8YFLogxK

U2 - 10.1117/12.2612893

DO - 10.1117/12.2612893

M3 - Conference contribution

AN - SCOPUS:85130012056

T3 - Progress in Biomedical Optics and Imaging - Proceedings of SPIE

BT - Medical Imaging 2022

A2 - Mello-Thoms, Claudia R.

A2 - Taylor-Phillips, Sian

PB - SPIE

T2 - Medical Imaging 2022: Image Perception, Observer Performance, and Technology Assessment

Y2 - 21 March 2022 through 27 March 2022

ER -

Evaluating Procedures for Establishing Generative Adversarial Network-based Stochastic Image Models in Medical Imaging

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this