Detection of Longitudinal Visual Field Progression in Glaucoma Using Machine Learning

Siamak Yousefi, Taichi Kiwaki, Yuhui Zheng, Hiroki Sugiura, Ryo Asaoka, Hiroshi Murata, Hans Lemij, Kenji Yamanishi

Research output: Contribution to journalArticle

6 Citations (Scopus)

Abstract

Purpose: Global indices of standard automated perimerty are insensitive to localized losses, while point-wise indices are sensitive but highly variable. Region-wise indices sit in between. This study introduces a machine learning–based index for glaucoma progression detection that outperforms global, region-wise, and point-wise indices. Design: Development and comparison of a prognostic index. Method: Visual fields from 2085 eyes of 1214 subjects were used to identify glaucoma progression patterns using machine learning. Visual fields from 133 eyes of 71 glaucoma patients were collected 10 times over 10 weeks to provide a no-change, test-retest dataset. The parameters of all methods were identified using visual field sequences in the test-retest dataset to meet fixed 95% specificity. An independent dataset of 270 eyes of 136 glaucoma patients and survival analysis were used to compare methods. Results: The time to detect progression in 25% of the eyes in the longitudinal dataset using global mean deviation (MD) was 5.2 (95% confidence interval, 4.1–6.5) years; 4.5 (4.0–5.5) years using region-wise, 3.9 (3.5–4.6) years using point-wise, and 3.5 (3.1–4.0) years using machine learning analysis. The time until 25% of eyes showed subsequently confirmed progression after 2 additional visits were included were 6.6 (5.6–7.4) years, 5.7 (4.8–6.7) years, 5.6 (4.7–6.5) years, and 5.1 (4.5–6.0) years for global, region-wise, point-wise, and machine learning analyses, respectively. Conclusions: Machine learning analysis detects progressing eyes earlier than other methods consistently, with or without confirmation visits. In particular, machine learning detects more slowly progressing eyes than other methods.

Original languageEnglish (US)
Pages (from-to)71-79
Number of pages9
JournalAmerican Journal of Ophthalmology
Volume193
DOIs
StatePublished - Sep 1 2018

Fingerprint

Visual Fields
Glaucoma
Survival Analysis
Machine Learning
Confidence Intervals
Datasets

All Science Journal Classification (ASJC) codes

  • Ophthalmology

Cite this

Detection of Longitudinal Visual Field Progression in Glaucoma Using Machine Learning. / Yousefi, Siamak; Kiwaki, Taichi; Zheng, Yuhui; Sugiura, Hiroki; Asaoka, Ryo; Murata, Hiroshi; Lemij, Hans; Yamanishi, Kenji.

In: American Journal of Ophthalmology, Vol. 193, 01.09.2018, p. 71-79.

Research output: Contribution to journalArticle

Yousefi, S, Kiwaki, T, Zheng, Y, Sugiura, H, Asaoka, R, Murata, H, Lemij, H & Yamanishi, K 2018, 'Detection of Longitudinal Visual Field Progression in Glaucoma Using Machine Learning', American Journal of Ophthalmology, vol. 193, pp. 71-79. https://doi.org/10.1016/j.ajo.2018.06.007
Yousefi, Siamak ; Kiwaki, Taichi ; Zheng, Yuhui ; Sugiura, Hiroki ; Asaoka, Ryo ; Murata, Hiroshi ; Lemij, Hans ; Yamanishi, Kenji. / Detection of Longitudinal Visual Field Progression in Glaucoma Using Machine Learning. In: American Journal of Ophthalmology. 2018 ; Vol. 193. pp. 71-79.
@article{58ba07f4106b4101bcc38c23659ef81b,
title = "Detection of Longitudinal Visual Field Progression in Glaucoma Using Machine Learning",
abstract = "Purpose: Global indices of standard automated perimerty are insensitive to localized losses, while point-wise indices are sensitive but highly variable. Region-wise indices sit in between. This study introduces a machine learning–based index for glaucoma progression detection that outperforms global, region-wise, and point-wise indices. Design: Development and comparison of a prognostic index. Method: Visual fields from 2085 eyes of 1214 subjects were used to identify glaucoma progression patterns using machine learning. Visual fields from 133 eyes of 71 glaucoma patients were collected 10 times over 10 weeks to provide a no-change, test-retest dataset. The parameters of all methods were identified using visual field sequences in the test-retest dataset to meet fixed 95{\%} specificity. An independent dataset of 270 eyes of 136 glaucoma patients and survival analysis were used to compare methods. Results: The time to detect progression in 25{\%} of the eyes in the longitudinal dataset using global mean deviation (MD) was 5.2 (95{\%} confidence interval, 4.1–6.5) years; 4.5 (4.0–5.5) years using region-wise, 3.9 (3.5–4.6) years using point-wise, and 3.5 (3.1–4.0) years using machine learning analysis. The time until 25{\%} of eyes showed subsequently confirmed progression after 2 additional visits were included were 6.6 (5.6–7.4) years, 5.7 (4.8–6.7) years, 5.6 (4.7–6.5) years, and 5.1 (4.5–6.0) years for global, region-wise, point-wise, and machine learning analyses, respectively. Conclusions: Machine learning analysis detects progressing eyes earlier than other methods consistently, with or without confirmation visits. In particular, machine learning detects more slowly progressing eyes than other methods.",
author = "Siamak Yousefi and Taichi Kiwaki and Yuhui Zheng and Hiroki Sugiura and Ryo Asaoka and Hiroshi Murata and Hans Lemij and Kenji Yamanishi",
year = "2018",
month = "9",
day = "1",
doi = "10.1016/j.ajo.2018.06.007",
language = "English (US)",
volume = "193",
pages = "71--79",
journal = "American Journal of Ophthalmology",
issn = "0002-9394",
publisher = "Elsevier USA",

}

TY - JOUR

T1 - Detection of Longitudinal Visual Field Progression in Glaucoma Using Machine Learning

AU - Yousefi, Siamak

AU - Kiwaki, Taichi

AU - Zheng, Yuhui

AU - Sugiura, Hiroki

AU - Asaoka, Ryo

AU - Murata, Hiroshi

AU - Lemij, Hans

AU - Yamanishi, Kenji

PY - 2018/9/1

Y1 - 2018/9/1

N2 - Purpose: Global indices of standard automated perimerty are insensitive to localized losses, while point-wise indices are sensitive but highly variable. Region-wise indices sit in between. This study introduces a machine learning–based index for glaucoma progression detection that outperforms global, region-wise, and point-wise indices. Design: Development and comparison of a prognostic index. Method: Visual fields from 2085 eyes of 1214 subjects were used to identify glaucoma progression patterns using machine learning. Visual fields from 133 eyes of 71 glaucoma patients were collected 10 times over 10 weeks to provide a no-change, test-retest dataset. The parameters of all methods were identified using visual field sequences in the test-retest dataset to meet fixed 95% specificity. An independent dataset of 270 eyes of 136 glaucoma patients and survival analysis were used to compare methods. Results: The time to detect progression in 25% of the eyes in the longitudinal dataset using global mean deviation (MD) was 5.2 (95% confidence interval, 4.1–6.5) years; 4.5 (4.0–5.5) years using region-wise, 3.9 (3.5–4.6) years using point-wise, and 3.5 (3.1–4.0) years using machine learning analysis. The time until 25% of eyes showed subsequently confirmed progression after 2 additional visits were included were 6.6 (5.6–7.4) years, 5.7 (4.8–6.7) years, 5.6 (4.7–6.5) years, and 5.1 (4.5–6.0) years for global, region-wise, point-wise, and machine learning analyses, respectively. Conclusions: Machine learning analysis detects progressing eyes earlier than other methods consistently, with or without confirmation visits. In particular, machine learning detects more slowly progressing eyes than other methods.

AB - Purpose: Global indices of standard automated perimerty are insensitive to localized losses, while point-wise indices are sensitive but highly variable. Region-wise indices sit in between. This study introduces a machine learning–based index for glaucoma progression detection that outperforms global, region-wise, and point-wise indices. Design: Development and comparison of a prognostic index. Method: Visual fields from 2085 eyes of 1214 subjects were used to identify glaucoma progression patterns using machine learning. Visual fields from 133 eyes of 71 glaucoma patients were collected 10 times over 10 weeks to provide a no-change, test-retest dataset. The parameters of all methods were identified using visual field sequences in the test-retest dataset to meet fixed 95% specificity. An independent dataset of 270 eyes of 136 glaucoma patients and survival analysis were used to compare methods. Results: The time to detect progression in 25% of the eyes in the longitudinal dataset using global mean deviation (MD) was 5.2 (95% confidence interval, 4.1–6.5) years; 4.5 (4.0–5.5) years using region-wise, 3.9 (3.5–4.6) years using point-wise, and 3.5 (3.1–4.0) years using machine learning analysis. The time until 25% of eyes showed subsequently confirmed progression after 2 additional visits were included were 6.6 (5.6–7.4) years, 5.7 (4.8–6.7) years, 5.6 (4.7–6.5) years, and 5.1 (4.5–6.0) years for global, region-wise, point-wise, and machine learning analyses, respectively. Conclusions: Machine learning analysis detects progressing eyes earlier than other methods consistently, with or without confirmation visits. In particular, machine learning detects more slowly progressing eyes than other methods.

UR - http://www.scopus.com/inward/record.url?scp=85049312694&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85049312694&partnerID=8YFLogxK

U2 - 10.1016/j.ajo.2018.06.007

DO - 10.1016/j.ajo.2018.06.007

M3 - Article

VL - 193

SP - 71

EP - 79

JO - American Journal of Ophthalmology

JF - American Journal of Ophthalmology

SN - 0002-9394

ER -