A deep learning system for detecting systemic lupus erythematosus from retinal images

Tingyao Li¹, Shiqun Lin², Zhouyu Guan³, Yukun Zhou⁴, Dian Zeng⁵, Zheyuan Wang⁶, Yan Zhou², Pinqi Fang⁷, Shujie Yu⁸, Ruhan Liu⁹, Xiang Chen⁶, Yan-Ran Joyce Wang¹⁰, Yuwei Lu¹, Jia Shu⁶, Yiming Qin¹¹, Yiting Wu⁷, Yilan Wu⁵, Chan Wu², Shangzhu Zhang¹², Jie Shen⁷, Huating Li³, Tingli Chen¹³, Jin Li¹⁴, Yih-Chung Tham¹⁵, Charumathi Sabanayagam¹⁶, Ying Feng Zheng¹⁷, Siegfried K Wagner¹⁸, Pearse A Keane¹⁸, Tien Yin Wong¹⁹, Rongping Dai²⁰, Bin Sheng²¹

Affiliations

¹ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; MOE Key Laboratory of AI, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China; Institute for Proactive Healthcare, Shanghai Jiao Tong University, Shanghai, China.
² Department of Ophthalmology, Peking Union Medical College Hospital, Peking Union Medical College, Chinese Academy of Medical Sciences, Beijing, China; Beijing Key Laboratory of Fundus Diseases Intelligent Diagnosis & Drug/Device Development and Translation, Beijing, China.
³ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; Institute for Proactive Healthcare, Shanghai Jiao Tong University, Shanghai, China.
⁴ Centre for Medical Image Computing, University College London, London, UK; NIHR Biomedical Research Centre at Moorfields Eye Hospital NHS Foundation Trust, London, UK; Institute of Ophthalmology, University College London, London, UK.
⁵ School of Clinical Medicine, Beijing Tsinghua Changgung Hospital, Tsinghua Medicine, Tsinghua University, Beijing, China.
⁶ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; MOE Key Laboratory of AI, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China.
⁷ Medical Records and Statistics Office, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China.
⁸ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China.
⁹ Furong Laboratory, Central South University, Changsha, Hunan, China.
¹⁰ School of Medicine, Stanford University, Stanford, CA, USA.
¹¹ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; MOE Key Laboratory of AI, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China; School of Clinical Medicine, Beijing Tsinghua Changgung Hospital, Tsinghua Medicine, Tsinghua University, Beijing, China.
¹² Department of Rheumatology and Clinical Immunology, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Center for Dermatologic and Immunologic Diseases (NCRC-DID), Ministry of Science & Technology, State Key Laboratory of Complex Severe and Rare Diseases, Peking Union Medical College Hospital (PUMCH), Key Laboratory of Rheumatology and Clinical Immunology, Ministry of Education, Beijing, China.
¹³ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; Department of Ophthalmology, Shanghai Health and Medical Center, Wuxi, China.
¹⁴ Department of Ophthalmology, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China.
¹⁵ Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore; Centre for Innovation and Precision Eye Health, and Department of Ophthalmology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore; Ophthalmology and Visual Science Academic Clinical Program, Duke-NUS Medical School, Singapore, Singapore.
¹⁶ Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore; Ophthalmology and Visual Science Academic Clinical Program, Duke-NUS Medical School, Singapore, Singapore.
¹⁷ Zhongshan Ophthalmic Center, Guangzhou, China.
¹⁸ NIHR Biomedical Research Centre at Moorfields Eye Hospital NHS Foundation Trust, London, UK; Institute of Ophthalmology, University College London, London, UK.
¹⁹ School of Clinical Medicine, Beijing Tsinghua Changgung Hospital, Tsinghua Medicine, Tsinghua University, Beijing, China; Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore; Zhongshan Ophthalmic Center, Guangzhou, China; Beijing Visual Science and Translational Eye Research Institute (BERI), Beijing Tsinghua Changgung Hospital, Beijing, China; Beijing Key Laboratory of Intelligent Diagnostic Technology and Devices for Major Blinding Eye Diseases, Tsinghua Medicine, Tsinghua University, Beijing, China.
²⁰ Department of Ophthalmology, Peking Union Medical College Hospital, Peking Union Medical College, Chinese Academy of Medical Sciences, Beijing, China; Beijing Key Laboratory of Fundus Diseases Intelligent Diagnosis & Drug/Device Development and Translation, Beijing, China. Electronic address: derricka@sina.com.
²¹ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; MOE Key Laboratory of AI, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China; Institute for Proactive Healthcare, Shanghai Jiao Tong University, Shanghai, China. Electronic address: shengbin@sjtu.edu.cn.

PMID: 40570853
PMCID: PMC12281353
DOI: 10.1016/j.xcrm.2025.102203

A deep learning system for detecting systemic lupus erythematosus from retinal images

Tingyao Li et al. Cell Rep Med. 2025.

. 2025 Jul 15;6(7):102203.

doi: 10.1016/j.xcrm.2025.102203. Epub 2025 Jun 25.

Authors

Affiliations

¹ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; MOE Key Laboratory of AI, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China; Institute for Proactive Healthcare, Shanghai Jiao Tong University, Shanghai, China.
² Department of Ophthalmology, Peking Union Medical College Hospital, Peking Union Medical College, Chinese Academy of Medical Sciences, Beijing, China; Beijing Key Laboratory of Fundus Diseases Intelligent Diagnosis & Drug/Device Development and Translation, Beijing, China.
³ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; Institute for Proactive Healthcare, Shanghai Jiao Tong University, Shanghai, China.
⁴ Centre for Medical Image Computing, University College London, London, UK; NIHR Biomedical Research Centre at Moorfields Eye Hospital NHS Foundation Trust, London, UK; Institute of Ophthalmology, University College London, London, UK.
⁵ School of Clinical Medicine, Beijing Tsinghua Changgung Hospital, Tsinghua Medicine, Tsinghua University, Beijing, China.
⁶ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; MOE Key Laboratory of AI, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China.
⁷ Medical Records and Statistics Office, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai, China.
⁸ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China.
⁹ Furong Laboratory, Central South University, Changsha, Hunan, China.
¹⁰ School of Medicine, Stanford University, Stanford, CA, USA.
¹¹ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; MOE Key Laboratory of AI, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China; School of Clinical Medicine, Beijing Tsinghua Changgung Hospital, Tsinghua Medicine, Tsinghua University, Beijing, China.
¹² Department of Rheumatology and Clinical Immunology, Chinese Academy of Medical Sciences & Peking Union Medical College, National Clinical Research Center for Dermatologic and Immunologic Diseases (NCRC-DID), Ministry of Science & Technology, State Key Laboratory of Complex Severe and Rare Diseases, Peking Union Medical College Hospital (PUMCH), Key Laboratory of Rheumatology and Clinical Immunology, Ministry of Education, Beijing, China.
¹³ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; Department of Ophthalmology, Shanghai Health and Medical Center, Wuxi, China.
¹⁴ Department of Ophthalmology, Renji Hospital, Shanghai Jiao Tong University School of Medicine, Shanghai, China.
¹⁵ Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore; Centre for Innovation and Precision Eye Health, and Department of Ophthalmology, Yong Loo Lin School of Medicine, National University of Singapore, Singapore, Singapore; Ophthalmology and Visual Science Academic Clinical Program, Duke-NUS Medical School, Singapore, Singapore.
¹⁶ Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore; Ophthalmology and Visual Science Academic Clinical Program, Duke-NUS Medical School, Singapore, Singapore.
¹⁷ Zhongshan Ophthalmic Center, Guangzhou, China.
¹⁸ NIHR Biomedical Research Centre at Moorfields Eye Hospital NHS Foundation Trust, London, UK; Institute of Ophthalmology, University College London, London, UK.
¹⁹ School of Clinical Medicine, Beijing Tsinghua Changgung Hospital, Tsinghua Medicine, Tsinghua University, Beijing, China; Singapore Eye Research Institute, Singapore National Eye Centre, Singapore, Singapore; Zhongshan Ophthalmic Center, Guangzhou, China; Beijing Visual Science and Translational Eye Research Institute (BERI), Beijing Tsinghua Changgung Hospital, Beijing, China; Beijing Key Laboratory of Intelligent Diagnostic Technology and Devices for Major Blinding Eye Diseases, Tsinghua Medicine, Tsinghua University, Beijing, China.
²⁰ Department of Ophthalmology, Peking Union Medical College Hospital, Peking Union Medical College, Chinese Academy of Medical Sciences, Beijing, China; Beijing Key Laboratory of Fundus Diseases Intelligent Diagnosis & Drug/Device Development and Translation, Beijing, China. Electronic address: derricka@sina.com.
²¹ Shanghai International Joint Laboratory of Intelligent Prevention and Treatment for Metabolic Diseases, Department of Computer Science and Engineering, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Department of Endocrinology and Metabolism, Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Shanghai Diabetes Institute, Shanghai Clinical Center for Diabetes, Shanghai, China; MOE Key Laboratory of AI, School of Electronic, Information, and Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China; Institute for Proactive Healthcare, Shanghai Jiao Tong University, Shanghai, China. Electronic address: shengbin@sjtu.edu.cn.

PMID: 40570853
PMCID: PMC12281353
DOI: 10.1016/j.xcrm.2025.102203

Abstract

Systemic lupus erythematosus (SLE) is a serious autoimmune disorder predominantly affecting women. However, screening for SLE and related complications poses significant challenges globally, due to complex diagnostic criteria and public unawareness. Since SLE-related retinal involvement could provide insights into disease activity and severity, we develop a deep learning system (DeepSLE) to detect SLE and its retinal and kidney complications from retinal images. In multi-ethnic validation datasets comprising 247,718 images from China and UK, DeepSLE achieves areas under the receiver operating characteristic curve of 0.822-0.969 for SLE. Additionally, DeepSLE demonstrates robust performance across subgroups stratified by gender, age, ethnicity, and socioeconomic status. To ensure DeepSLE's explainability, we conduct both qualitative and quantitative analyses. Furthermore, in a prospective reader study, DeepSLE demonstrates higher sensitivities compared with primary care physicians. Altogether, DeepSLE offers digital solutions for detecting SLE and related complications from retinal images, holding potential for future clinical deployment.

Keywords: color fundus photography; deep learning; health equity; lupus nephritis; lupus retinopathy; multi-ethnic validation; opportunistic screening; reader study; retinal imaging; systemic lupus erythematosus.

PubMed Disclaimer

Conflict of interest statement

Declaration of interests The authors declare no competing interests.

Figures

**Figure 1**
Overview study design of the DeepSLE system (A) Graphical illustration of the DeepSLE system. The DeepSLE system could conduct three clinical tasks using retinal fundus images as inputs, including the detection of systemic lupus erythematosus (SLE), lupus retinopathy (LR), and lupus nephritis (LN). The DeepSLE system was first pre-trained in a self-supervised learning way to extract transferable visual representations from retinal fundus images and then adapted to these three clinical tasks. (B) Retrospective multi-ethnic evaluations of the DeepSLE system in the internal and external validation datasets. Four external validation datasets from China and the UK were included. (C) Subgroup analyses of the DeepSLE system for SLE detection. To ensure the fairness of the DeepSLE system, we conducted analyses on a wide range of patient subgroups, with respect to gender and age in the internal validation set and ethnicity and socioeconomic status in external test sets. (D) Explainability analysis. To better understand how the DeepSLE system could detect SLE, LR, and LN, we conducted both qualitative and quantitative analyses to ensure the relevance and interpretability of the resulting features. (E) Prospective reader study. We conducted a prospective reader study to compare the performance of the DeepSLE system with that of physicians for detecting SLE, LR, and LN in primary care settings. Five primary care physicians and five immunology specialists were recruited. Figure 1 was created with https://BioRender.com.

**Figure 2**
Performance of the DeepSLE system on validation sets and among different subgroups Each subgroup was evaluated using three metrics: AUROC, sensitivity, and specificity. Green (SLE cases) and purple (non-SLE cases) bars represent the number of patients in each subgroup. Metrics are for all subgroups and are reported with 95% CIs calculated by bootstrapping with 1,000 replicates. PUMCH, Peking Union Medical College Hospital; SSPH, Shanghai Six People’s Hospital dataset; SDPP, Shanghai Diabetes Prevention Program dataset; UKB, the United Kingdom Biobank; MEH, the Moorfields Eye Hospital dataset; SLE, systemic lupus erythematosus. (A) Performance of the DeepSLE system on validation sets. Receiver operating characteristic (ROC) curves with area under the receiver operating characteristic curve (AUROC) of the DeepSLE system were shown for detecting SLE in the internal validation set and external validation sets. (B) System performance for detecting SLE across demographic subgroups stratified by gender and age on the internal test set. Participants were categorized as under 18 years (subgroup 1), 18–45 years (subgroup 2), or over 45 years (subgroup 3). (C) System performance for detecting SLE across demographic subgroups stratified by ethnicity and socioeconomic status on the UKB dataset. The socioeconomic status was measured by Townsend deprivation index deciles. Participants were categorized as least deprived (decile 1 to decile 3), moderately deprived (decile 4 to decile 7), or most deprived (decile 8 to decile 10). (D) System performance for detecting SLE across demographic subgroups stratified by ethnicity and socioeconomic status on the MEH dataset. The socioeconomic status was measured by the Index of Multiple Deprivation deciles. Participants were categorized as most deprived (decile 1 to decile 3), moderately deprived (decile 4 to decile 7), and least deprived (decile 8 to decile 10).

**Figure 3**
Explainability analyses of the DeepSLE system (A) Qualitative analysis using saliency maps. The results showed that our DeepSLE system focused on the retinal vessels, the macula, and retinal lesions for disease detections. (B) Quantitative analysis of retinal vascular variables for color fundus photographs with SLE and without SLE. First, we performed vessel segmentation on CFPs in various regions to get vascular contours. Using the segmented images, we quantified a range of retinal vascular variables, including fractal dimension related to vessel complexity, distance tortuosity and squared curvature tortuosity related to vessel tortuosity, and central retinal arteriolar equivalent (CRAE) and central retinal venular equivalent (CRVE) related to vessel caliber. p values for comparing the retinal vascular variables between CFPs with and without SLE were shown. p values were calculated using the Mann-Whitney U test.

**Figure 4**
Comparison of DeepSLE with physicians in a prospective reader study (A) Reader study design. Five primary care physicians and five immunology specialists were recruited to make diagnosis of SLE, LR, and LN, based on age, gender, medical history, findings from physical examinations, and retinal fundus images. The AI model (DeepSLE) was provided with only the retinal fundus photographs (created with https://BioRender.com). (B) Comparison of DeepSLE’s performance with that of PCPs and specialists for detecting SLE. The sensitivity of DeepSLE outperformed all the PCPs for detecting SLE, while no significant differences in specificity were observed between DeepSLE and the PCPs.

See this image and copyright information in PMC

References

1. Kaul A., Gordon C., Crow M.K., Touma Z., Urowitz M.B., Van Vollenhoven R., Ruiz-Irastorza G., Hughes G. Systemic lupus erythematosus. Nat. Rev. Dis. Primers. 2016;2:16039. - PubMed
1. Gómez-Bañuelos E., Goldman D.W., Andrade V., Darrah E., Petri M., Andrade F. Uncoupling interferons and the interferon signature explains clinical and transcriptional subsets in SLE. Cell Rep. Med. 2024;5:101569. doi: 10.1016/j.xcrm.2024.101569. - DOI - PMC - PubMed
1. Kain J., Owen K.A., Marion M.C., Langefeld C.D., Grammer A.C., Lipsky P.E. Mendelian randomization and pathway analysis demonstrate shared genetic associations between lupus and coronary artery disease. Cell Rep. Med. 2022;3 doi: 10.1016/j.xcrm.2022.100805. - DOI - PMC - PubMed
1. Jiang S.H., Mercan S., Papa I., Moldovan M., Walters G.D., Koina M., Fadia M., Stanley M., Lea-Henry T., Cook A., et al. Deletions in VANGL1 are a risk factor for antibody-mediated kidney disease. Cell Rep. Med. 2021;2 doi: 10.1016/j.xcrm.2021.100475. - DOI - PMC - PubMed
1. Siegel C.H., Sammaritano L.R. Systemic Lupus Erythematosus: A Review. JAMA. 2024;331:1480–1491. doi: 10.1001/jama.2024.2315. - DOI - PubMed

MeSH terms

Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions
Actions

LinkOut - more resources

Full Text Sources
Medical
- MedlinePlus Health Information

Save citation to file

Email citation

Add to Collections

Add to My Bibliography

Your saved search

Create a file for external citation management software

Your RSS Feed

A deep learning system for detecting systemic lupus erythematosus from retinal images

Affiliations

A deep learning system for detecting systemic lupus erythematosus from retinal images

Authors

Affiliations

Abstract

Conflict of interest statement

Figures

References

MeSH terms

LinkOut - more resources

Full Text Sources

Medical