3D-CNN for glaucoma detection using optical coherence tomography

Yasmeen George; Bhavna Antony; Hiroshi Ishikawa; Gadi Wollstein; Joel Schuman; Rahil Garnavi

doi:10.1007/978-3-030-32956-3_7

3D-CNN for glaucoma detection using optical coherence tomography

Yasmeen George, Bhavna Antony, Hiroshi Ishikawa, Gadi Wollstein, Joel Schuman, Rahil Garnavi

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

3 Citations (Scopus)

Abstract

The large size of raw 3D optical coherence tomography (OCT) volumes poses challenges for deep learning methods as it cannot be accommodated on a single GPU in its original resolution. The direct analysis of these volumes however, provides advantages such as circumventing the need for the segmentation of retinal structures. Previously, a deep learning (DL) approach was proposed for the detection of glaucoma directly from 3D OCT volumes, where the volumes were significantly downsampled first. In this paper, we propose an end-to-end DL model for the detection of glaucoma that doubles the number of input voxels of the previously proposed method, and also boasts an improved AUC = 0.973 over the results obtained using the previously proposed approach of AUC = 0.946. Furthermore, this paper also includes a quantitative analysis of the regions of the volume highlighted by grad-CAM visualization. Occlusion of these highlighted regions resulted in a drop in performance by 40%, indicating that the regions highlighted by gradient-weighted class activation maps (grad-CAM) are indeed crucial to the performance of the model.

Original language	English
Title of host publication	6th International Workshop, OMIA 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 17 Proceedings
Editors	Huazhu Fu, Mona K. Garvin, Tom MacGillivray, Yanwu Xu, Yalin Zheng
Place of Publication	Cham Switzerland
Publisher	Springer
Pages	52-59
Number of pages	8
ISBN (Electronic)	9783030329563
ISBN (Print)	9783030329556
DOIs	https://doi.org/10.1007/978-3-030-32956-3_7
Publication status	Published - 2019
Externally published	Yes
Event	International Workshop on Ophthalmic Medical Image Analysis 2019 - Shenzhen, China Duration: 17 Oct 2019 → 17 Oct 2019 Conference number: 6th https://link.springer.com/book/10.1007/978-3-030-32956-3 (Proceedings)

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	11855
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	International Workshop on Ophthalmic Medical Image Analysis 2019
Abbreviated title	OMIA 2019
Country/Territory	China
City	Shenzhen
Period	17/10/19 → 17/10/19
Internet address	https://link.springer.com/book/10.1007/978-3-030-32956-3 (Proceedings)

Keywords

3D-CNN
Glaucoma detection
Gradient-weighted class activation maps
Optical coherence tomography
Visual explanations

Access to Document

10.1007/978-3-030-32956-3_7

Cite this

George, Y., Antony, B., Ishikawa, H., Wollstein, G., Schuman, J., & Garnavi, R. (2019). 3D-CNN for glaucoma detection using optical coherence tomography. In H. Fu, M. K. Garvin, T. MacGillivray, Y. Xu, & Y. Zheng (Eds.), 6th International Workshop, OMIA 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 17 Proceedings (pp. 52-59). (Lecture Notes in Computer Science; Vol. 11855). Springer. https://doi.org/10.1007/978-3-030-32956-3_7

George, Yasmeen ; Antony, Bhavna ; Ishikawa, Hiroshi et al. / 3D-CNN for glaucoma detection using optical coherence tomography. 6th International Workshop, OMIA 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 17 Proceedings. editor / Huazhu Fu ; Mona K. Garvin ; Tom MacGillivray ; Yanwu Xu ; Yalin Zheng. Cham Switzerland : Springer, 2019. pp. 52-59 (Lecture Notes in Computer Science).

@inproceedings{0e8721d0d3c347eba01d9b4e246f0992,

title = "3D-CNN for glaucoma detection using optical coherence tomography",

abstract = "The large size of raw 3D optical coherence tomography (OCT) volumes poses challenges for deep learning methods as it cannot be accommodated on a single GPU in its original resolution. The direct analysis of these volumes however, provides advantages such as circumventing the need for the segmentation of retinal structures. Previously, a deep learning (DL) approach was proposed for the detection of glaucoma directly from 3D OCT volumes, where the volumes were significantly downsampled first. In this paper, we propose an end-to-end DL model for the detection of glaucoma that doubles the number of input voxels of the previously proposed method, and also boasts an improved AUC = 0.973 over the results obtained using the previously proposed approach of AUC = 0.946. Furthermore, this paper also includes a quantitative analysis of the regions of the volume highlighted by grad-CAM visualization. Occlusion of these highlighted regions resulted in a drop in performance by 40%, indicating that the regions highlighted by gradient-weighted class activation maps (grad-CAM) are indeed crucial to the performance of the model.",

keywords = "3D-CNN, Glaucoma detection, Gradient-weighted class activation maps, Optical coherence tomography, Visual explanations",

author = "Yasmeen George and Bhavna Antony and Hiroshi Ishikawa and Gadi Wollstein and Joel Schuman and Rahil Garnavi",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2019.; International Workshop on Ophthalmic Medical Image Analysis 2019, OMIA 2019 ; Conference date: 17-10-2019 Through 17-10-2019",

year = "2019",

doi = "10.1007/978-3-030-32956-3_7",

language = "English",

isbn = "9783030329556",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "52--59",

editor = "Huazhu Fu and Garvin, {Mona K.} and Tom MacGillivray and Yanwu Xu and Yalin Zheng",

booktitle = "6th International Workshop, OMIA 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 17 Proceedings",

url = "https://link.springer.com/book/10.1007/978-3-030-32956-3",

}

George, Y, Antony, B, Ishikawa, H, Wollstein, G, Schuman, J & Garnavi, R 2019, 3D-CNN for glaucoma detection using optical coherence tomography. in H Fu, MK Garvin, T MacGillivray, Y Xu & Y Zheng (eds), 6th International Workshop, OMIA 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 17 Proceedings. Lecture Notes in Computer Science, vol. 11855, Springer, Cham Switzerland, pp. 52-59, International Workshop on Ophthalmic Medical Image Analysis 2019, Shenzhen, China, 17/10/19. https://doi.org/10.1007/978-3-030-32956-3_7

3D-CNN for glaucoma detection using optical coherence tomography. / George, Yasmeen; Antony, Bhavna; Ishikawa, Hiroshi et al.
6th International Workshop, OMIA 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 17 Proceedings. ed. / Huazhu Fu; Mona K. Garvin; Tom MacGillivray; Yanwu Xu; Yalin Zheng. Cham Switzerland: Springer, 2019. p. 52-59 (Lecture Notes in Computer Science; Vol. 11855).

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

TY - GEN

T1 - 3D-CNN for glaucoma detection using optical coherence tomography

AU - George, Yasmeen

AU - Antony, Bhavna

AU - Ishikawa, Hiroshi

AU - Wollstein, Gadi

AU - Schuman, Joel

AU - Garnavi, Rahil

N1 - Conference code: 6th

PY - 2019

Y1 - 2019

N2 - The large size of raw 3D optical coherence tomography (OCT) volumes poses challenges for deep learning methods as it cannot be accommodated on a single GPU in its original resolution. The direct analysis of these volumes however, provides advantages such as circumventing the need for the segmentation of retinal structures. Previously, a deep learning (DL) approach was proposed for the detection of glaucoma directly from 3D OCT volumes, where the volumes were significantly downsampled first. In this paper, we propose an end-to-end DL model for the detection of glaucoma that doubles the number of input voxels of the previously proposed method, and also boasts an improved AUC = 0.973 over the results obtained using the previously proposed approach of AUC = 0.946. Furthermore, this paper also includes a quantitative analysis of the regions of the volume highlighted by grad-CAM visualization. Occlusion of these highlighted regions resulted in a drop in performance by 40%, indicating that the regions highlighted by gradient-weighted class activation maps (grad-CAM) are indeed crucial to the performance of the model.

AB - The large size of raw 3D optical coherence tomography (OCT) volumes poses challenges for deep learning methods as it cannot be accommodated on a single GPU in its original resolution. The direct analysis of these volumes however, provides advantages such as circumventing the need for the segmentation of retinal structures. Previously, a deep learning (DL) approach was proposed for the detection of glaucoma directly from 3D OCT volumes, where the volumes were significantly downsampled first. In this paper, we propose an end-to-end DL model for the detection of glaucoma that doubles the number of input voxels of the previously proposed method, and also boasts an improved AUC = 0.973 over the results obtained using the previously proposed approach of AUC = 0.946. Furthermore, this paper also includes a quantitative analysis of the regions of the volume highlighted by grad-CAM visualization. Occlusion of these highlighted regions resulted in a drop in performance by 40%, indicating that the regions highlighted by gradient-weighted class activation maps (grad-CAM) are indeed crucial to the performance of the model.

KW - 3D-CNN

KW - Glaucoma detection

KW - Gradient-weighted class activation maps

KW - Optical coherence tomography

KW - Visual explanations

UR - http://www.scopus.com/inward/record.url?scp=85075660036&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-32956-3_7

DO - 10.1007/978-3-030-32956-3_7

M3 - Conference Paper

AN - SCOPUS:85075660036

SN - 9783030329556

T3 - Lecture Notes in Computer Science

SP - 52

EP - 59

BT - 6th International Workshop, OMIA 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 17 Proceedings

A2 - Fu, Huazhu

A2 - Garvin, Mona K.

A2 - MacGillivray, Tom

A2 - Xu, Yanwu

A2 - Zheng, Yalin

PB - Springer

CY - Cham Switzerland

T2 - International Workshop on Ophthalmic Medical Image Analysis 2019

Y2 - 17 October 2019 through 17 October 2019

ER -

George Y, Antony B, Ishikawa H, Wollstein G, Schuman J, Garnavi R. 3D-CNN for glaucoma detection using optical coherence tomography. In Fu H, Garvin MK, MacGillivray T, Xu Y, Zheng Y, editors, 6th International Workshop, OMIA 2019 Held in Conjunction with MICCAI 2019 Shenzhen, China, October 17 Proceedings. Cham Switzerland: Springer. 2019. p. 52-59. (Lecture Notes in Computer Science). doi: 10.1007/978-3-030-32956-3_7

3D-CNN for glaucoma detection using optical coherence tomography

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Cite this