Deep learning algorithms to detect diabetic kidney disease from retinal photographs in multiethnic populations with diabetes

Bjorn Kaijun Betzler; Evelyn Yi Lyn Chee; Feng He; Cynthia Ciwei Lim; Jinyi Ho; Haslina Hamzah; Ngiap Chuan Tan; Gerald Liew; Gareth J McKay; Ruth E Hogg; Ian S Young; Ching-Yu Cheng; Su Chi Lim; Aaron Y Lee; Tien Yin Wong; Mong Li Lee; Wynne Hsu; Gavin Siew Wei Tan; Charumathi Sabanayagam

doi:10.1093/jamia/ocad179

Deep learning algorithms to detect diabetic kidney disease from retinal photographs in multiethnic populations with diabetes

Bjorn Kaijun Betzler, Evelyn Yi Lyn Chee, Feng He, Cynthia Ciwei Lim, Jinyi Ho, Haslina Hamzah, Ngiap Chuan Tan, Gerald Liew, Gareth J McKay, Ruth E Hogg, Ian S Young, Ching-Yu Cheng, Su Chi Lim, Aaron Y Lee, Tien Yin Wong, Mong Li Lee, Wynne Hsu, Gavin Siew Wei Tan, Charumathi Sabanayagam

Research output: Contribution to journal › Article › peer-review

8 Citations (Scopus)

Abstract

OBJECTIVE: To develop a deep learning algorithm (DLA) to detect diabetic kideny disease (DKD) from retinal photographs of patients with diabetes, and evaluate performance in multiethnic populations.

MATERIALS AND METHODS: We trained 3 models: (1) image-only; (2) risk factor (RF)-only multivariable logistic regression (LR) model adjusted for age, sex, ethnicity, diabetes duration, HbA1c, systolic blood pressure; (3) hybrid multivariable LR model combining RF data and standardized z-scores from image-only model. Data from Singapore Integrated Diabetic Retinopathy Program (SiDRP) were used to develop (6066 participants with diabetes, primary-care-based) and internally validate (5-fold cross-validation) the models. External testing on 2 independent datasets: (1) Singapore Epidemiology of Eye Diseases (SEED) study (1885 participants with diabetes, population-based); (2) Singapore Macroangiopathy and Microvascular Reactivity in Type 2 Diabetes (SMART2D) (439 participants with diabetes, cross-sectional) in Singapore. Supplementary external testing on 2 Caucasian cohorts: (3) Australian Eye and Heart Study (AHES) (460 participants with diabetes, cross-sectional) and (4) Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA) (265 participants with diabetes, cross-sectional).

RESULTS: In SiDRP validation, area under the curve (AUC) was 0.826(95% CI 0.818-0.833) for image-only, 0.847(0.840-0.854) for RF-only, and 0.866(0.859-0.872) for hybrid. Estimates with SEED were 0.764(0.743-0.785) for image-only, 0.802(0.783-0.822) for RF-only, and 0.828(0.810-0.846) for hybrid. In SMART2D, AUC was 0.726(0.686-0.765) for image-only, 0.701(0.660-0.741) in RF-only, 0.761(0.724-0.797) for hybrid.

DISCUSSION AND CONCLUSION: There is potential for DLA using retinal images as a screening adjunct for DKD among individuals with diabetes. This can value-add to existing DLA systems which diagnose diabetic retinopathy from retinal images, facilitating primary screening for DKD.

Original language	English
Journal	Journal of the American Medical Informatics Association
Early online date	02 Sept 2023
DOIs	https://doi.org/10.1093/jamia/ocad179
Publication status	Early online date - 02 Sept 2023

Bibliographical note

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1093/jamia/ocad179Licence: CC BY

Cite this

Betzler, B. K., Chee, E. Y. L., He, F., Lim, C. C., Ho, J., Hamzah, H., Tan, N. C., Liew, G., McKay, G. J., Hogg, R. E., Young, I. S., Cheng, C.-Y., Lim, S. C., Lee, A. Y., Wong, T. Y., Lee, M. L., Hsu, W., Tan, G. S. W., & Sabanayagam, C. (2023). Deep learning algorithms to detect diabetic kidney disease from retinal photographs in multiethnic populations with diabetes. Journal of the American Medical Informatics Association. Advance online publication. https://doi.org/10.1093/jamia/ocad179

@article{cac357311fcf4f53854b22d0bf839ec7,

title = "Deep learning algorithms to detect diabetic kidney disease from retinal photographs in multiethnic populations with diabetes",

abstract = "OBJECTIVE: To develop a deep learning algorithm (DLA) to detect diabetic kideny disease (DKD) from retinal photographs of patients with diabetes, and evaluate performance in multiethnic populations.MATERIALS AND METHODS: We trained 3 models: (1) image-only; (2) risk factor (RF)-only multivariable logistic regression (LR) model adjusted for age, sex, ethnicity, diabetes duration, HbA1c, systolic blood pressure; (3) hybrid multivariable LR model combining RF data and standardized z-scores from image-only model. Data from Singapore Integrated Diabetic Retinopathy Program (SiDRP) were used to develop (6066 participants with diabetes, primary-care-based) and internally validate (5-fold cross-validation) the models. External testing on 2 independent datasets: (1) Singapore Epidemiology of Eye Diseases (SEED) study (1885 participants with diabetes, population-based); (2) Singapore Macroangiopathy and Microvascular Reactivity in Type 2 Diabetes (SMART2D) (439 participants with diabetes, cross-sectional) in Singapore. Supplementary external testing on 2 Caucasian cohorts: (3) Australian Eye and Heart Study (AHES) (460 participants with diabetes, cross-sectional) and (4) Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA) (265 participants with diabetes, cross-sectional).RESULTS: In SiDRP validation, area under the curve (AUC) was 0.826(95% CI 0.818-0.833) for image-only, 0.847(0.840-0.854) for RF-only, and 0.866(0.859-0.872) for hybrid. Estimates with SEED were 0.764(0.743-0.785) for image-only, 0.802(0.783-0.822) for RF-only, and 0.828(0.810-0.846) for hybrid. In SMART2D, AUC was 0.726(0.686-0.765) for image-only, 0.701(0.660-0.741) in RF-only, 0.761(0.724-0.797) for hybrid.DISCUSSION AND CONCLUSION: There is potential for DLA using retinal images as a screening adjunct for DKD among individuals with diabetes. This can value-add to existing DLA systems which diagnose diabetic retinopathy from retinal images, facilitating primary screening for DKD.",

author = "Betzler, {Bjorn Kaijun} and Chee, {Evelyn Yi Lyn} and Feng He and Lim, {Cynthia Ciwei} and Jinyi Ho and Haslina Hamzah and Tan, {Ngiap Chuan} and Gerald Liew and McKay, {Gareth J} and Hogg, {Ruth E} and Young, {Ian S} and Ching-Yu Cheng and Lim, {Su Chi} and Lee, {Aaron Y} and Wong, {Tien Yin} and Lee, {Mong Li} and Wynne Hsu and Tan, {Gavin Siew Wei} and Charumathi Sabanayagam",

note = "{\textcopyright} The Author(s) 2023. Published by Oxford University Press on behalf of the American Medical Informatics Association.",

year = "2023",

month = sep,

day = "2",

doi = "10.1093/jamia/ocad179",

language = "English",

journal = "Journal of the American Medical Informatics Association",

issn = "1067-5027",

publisher = "Oxford University Press",

}

Betzler, BK, Chee, EYL, He, F, Lim, CC, Ho, J, Hamzah, H, Tan, NC, Liew, G, McKay, GJ , Hogg, RE , Young, IS, Cheng, C-Y, Lim, SC, Lee, AY, Wong, TY, Lee, ML, Hsu, W, Tan, GSW & Sabanayagam, C 2023, 'Deep learning algorithms to detect diabetic kidney disease from retinal photographs in multiethnic populations with diabetes', Journal of the American Medical Informatics Association. https://doi.org/10.1093/jamia/ocad179

TY - JOUR

T1 - Deep learning algorithms to detect diabetic kidney disease from retinal photographs in multiethnic populations with diabetes

AU - Betzler, Bjorn Kaijun

AU - Chee, Evelyn Yi Lyn

AU - He, Feng

AU - Lim, Cynthia Ciwei

AU - Ho, Jinyi

AU - Hamzah, Haslina

AU - Tan, Ngiap Chuan

AU - Liew, Gerald

AU - McKay, Gareth J

AU - Hogg, Ruth E

AU - Young, Ian S

AU - Cheng, Ching-Yu

AU - Lim, Su Chi

AU - Lee, Aaron Y

AU - Wong, Tien Yin

AU - Lee, Mong Li

AU - Hsu, Wynne

AU - Tan, Gavin Siew Wei

AU - Sabanayagam, Charumathi

N1 - © The Author(s) 2023. Published by Oxford University Press on behalf of the American Medical Informatics Association.

PY - 2023/9/2

Y1 - 2023/9/2

N2 - OBJECTIVE: To develop a deep learning algorithm (DLA) to detect diabetic kideny disease (DKD) from retinal photographs of patients with diabetes, and evaluate performance in multiethnic populations.MATERIALS AND METHODS: We trained 3 models: (1) image-only; (2) risk factor (RF)-only multivariable logistic regression (LR) model adjusted for age, sex, ethnicity, diabetes duration, HbA1c, systolic blood pressure; (3) hybrid multivariable LR model combining RF data and standardized z-scores from image-only model. Data from Singapore Integrated Diabetic Retinopathy Program (SiDRP) were used to develop (6066 participants with diabetes, primary-care-based) and internally validate (5-fold cross-validation) the models. External testing on 2 independent datasets: (1) Singapore Epidemiology of Eye Diseases (SEED) study (1885 participants with diabetes, population-based); (2) Singapore Macroangiopathy and Microvascular Reactivity in Type 2 Diabetes (SMART2D) (439 participants with diabetes, cross-sectional) in Singapore. Supplementary external testing on 2 Caucasian cohorts: (3) Australian Eye and Heart Study (AHES) (460 participants with diabetes, cross-sectional) and (4) Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA) (265 participants with diabetes, cross-sectional).RESULTS: In SiDRP validation, area under the curve (AUC) was 0.826(95% CI 0.818-0.833) for image-only, 0.847(0.840-0.854) for RF-only, and 0.866(0.859-0.872) for hybrid. Estimates with SEED were 0.764(0.743-0.785) for image-only, 0.802(0.783-0.822) for RF-only, and 0.828(0.810-0.846) for hybrid. In SMART2D, AUC was 0.726(0.686-0.765) for image-only, 0.701(0.660-0.741) in RF-only, 0.761(0.724-0.797) for hybrid.DISCUSSION AND CONCLUSION: There is potential for DLA using retinal images as a screening adjunct for DKD among individuals with diabetes. This can value-add to existing DLA systems which diagnose diabetic retinopathy from retinal images, facilitating primary screening for DKD.

AB - OBJECTIVE: To develop a deep learning algorithm (DLA) to detect diabetic kideny disease (DKD) from retinal photographs of patients with diabetes, and evaluate performance in multiethnic populations.MATERIALS AND METHODS: We trained 3 models: (1) image-only; (2) risk factor (RF)-only multivariable logistic regression (LR) model adjusted for age, sex, ethnicity, diabetes duration, HbA1c, systolic blood pressure; (3) hybrid multivariable LR model combining RF data and standardized z-scores from image-only model. Data from Singapore Integrated Diabetic Retinopathy Program (SiDRP) were used to develop (6066 participants with diabetes, primary-care-based) and internally validate (5-fold cross-validation) the models. External testing on 2 independent datasets: (1) Singapore Epidemiology of Eye Diseases (SEED) study (1885 participants with diabetes, population-based); (2) Singapore Macroangiopathy and Microvascular Reactivity in Type 2 Diabetes (SMART2D) (439 participants with diabetes, cross-sectional) in Singapore. Supplementary external testing on 2 Caucasian cohorts: (3) Australian Eye and Heart Study (AHES) (460 participants with diabetes, cross-sectional) and (4) Northern Ireland Cohort for the Longitudinal Study of Ageing (NICOLA) (265 participants with diabetes, cross-sectional).RESULTS: In SiDRP validation, area under the curve (AUC) was 0.826(95% CI 0.818-0.833) for image-only, 0.847(0.840-0.854) for RF-only, and 0.866(0.859-0.872) for hybrid. Estimates with SEED were 0.764(0.743-0.785) for image-only, 0.802(0.783-0.822) for RF-only, and 0.828(0.810-0.846) for hybrid. In SMART2D, AUC was 0.726(0.686-0.765) for image-only, 0.701(0.660-0.741) in RF-only, 0.761(0.724-0.797) for hybrid.DISCUSSION AND CONCLUSION: There is potential for DLA using retinal images as a screening adjunct for DKD among individuals with diabetes. This can value-add to existing DLA systems which diagnose diabetic retinopathy from retinal images, facilitating primary screening for DKD.

U2 - 10.1093/jamia/ocad179

DO - 10.1093/jamia/ocad179

M3 - Article

C2 - 37659103

SN - 1067-5027

JO - Journal of the American Medical Informatics Association

JF - Journal of the American Medical Informatics Association

ER -