AI analysis of retinal images: a novel non-invasive tool for diabetic nephropathy

Over 400 million people worldwide live with type 2 diabetes, and nearly half of these patients will develop chronic kidney disease over time. Diabetic nephropathy (DN) ranks as the leading trigger of irreversible end-stage renal disease (ESRD), imposing heavy medical burdens on healthcare systems globally. At present, the gold standard to distinguish DN from non-diabetic renal disease (NDRD) is invasive renal biopsy. However, this procedure brings unavoidable risks such as bleeding and secondary infection. It is also unavailable in remote and resource-limited medical facilities. Without pathological evidence, clinicians often make misdiagnoses based merely on routine blood and urine tests. Misclassification holds back precise targeted treatment, distorts disease prevalence data, and wastes limited medical resources, which creates a pressing demand for safe, convenient, and scalable non-invasive diagnostic solutions.

The retinal microvasculature is the only microcirculation in human bodies that can be directly photographed and observed, and its morphological abnormalities can accurately mirror glomerular microvascular damage induced by long-term diabetes. Previous studies only evaluated individual retinal vascular indicators in isolation, ignoring the internal correlation and synergistic mixed effects of multiple parameters. To address this research deficiency, an interdisciplinary team covering nephrology and ophthalmology from the Chinese PLA General Hospital and multiple cooperative medical institutions constructed an integrated machine learning analytical framework. Through weighted quantile sum (WQS) regression, the research team integrated all retinal vascular geometric indicators into a single composite mixed-effect biomarker for comprehensive evaluation.

Co-led by Dr. Sai Pan, Professor Xiaoniao Chen, Dr. Qian Wang, Dr. Guangyan Cai, and Professor Zheyi Dong, this multicenter research enrolled a total of 397 type 2 diabetic patients with biopsy-confirmed kidney lesions from six hospitals across China. The research population was split into a primary cohort of 374 participants and an independent external validation cohort containing 23 subjects. Researchers collected standard fundus photographs of all participants, adopted unsupervised learning algorithms paired with ResNet deep neural networks to realize automatic, labor-free retinal vessel segmentation, and calculated a full set of core geometric metrics, including vessel diameter, fractal dimension, tortuosity density, and vascular branching angles.

Seven mainstream machine learning algorithms were compared to screen the optimal model for DN differential diagnosis. The top-performing logistic regression model integrated the WQS retinal composite index and seven vital clinical indicators. It achieved an AUC of 0.98 and an accuracy of 0.92 on internal test datasets, and maintained robust performance with an AUC of 0.95 and accuracy of 0.91 in cross-hospital external validation. Ablation experiments further proved that incorporating retinal vascular parameters significantly improves overall diagnostic capacity.

To predict long-term renal outcomes, the team built a COX-Random Forest prognostic model based on 139 DN patients followed continuously until December 2023, with all-cause death or progression to ESRD as critical clinical endpoints. The model reached an AUC of 0.88, effectively identifying patients at high risk of rapid renal function decline. SHAP interpretability analysis identified diabetes duration, diabetic retinopathy, and age as the three most influential predictive factors.

Prof. Chen explained that the WQS composite index maximizes the diagnostic potential of routine fundus images, while the unsupervised segmentation technology supports fully automatic image interpretation even in grassroots community hospitals, demonstrating the great clinical value of fundus photographs in kidney disease diagnosis and management. First author Sai Pan stated that this multimodal AI pipeline breaks the limitation of single-index analysis and balances diagnostic accuracy and clinical accessibility, fully reflecting the unique advantages of AI in integrating multi-dimensional medical data for disease screening. Prof. Dong added that the stable multicenter validation results verify the model’s reliability for large-scale clinical screening, laying a solid foundation for its wide popularization and practical deployment in routine clinical practice in the future.

***

Reference

Title of original paper: Merging artificial intelligence and retinal vascular geometric parameters: A novel tool for diagnosis and prognosis prediction of diabetic nephropathy

Journal: Chinese Medical Journal

DOI: 10.1097/CM9.0000000000004211

About Xiaoniao Chen from Third Medical Center of Chinese People’s Liberation Army General Hospital
During the pursuit of a doctoral degree, the candidate received joint training at Harvard Medical School. For years, he/she has engaged in basic and clinical research focusing on corneal diseases and ocular trauma. His/her doctoral dissertation was awarded the title of “Excellent Doctoral Dissertation of the Chinese People’s Liberation Army.” As the first or corresponding author, he/she has published 20 papers in authoritative SCI journals, including top‑tier and prestigious ophthalmology‑related journals such as Signal Transduction and Targeted Therapy, The Ocular Surface, Investigative Ophthalmology & Visual Science, Biomaterials, Stem Cells International, and American Journal of Pathology, with a maximum impact factor of 52.7.

Funding information
This work was supported by grants from Science Technology Project of Beijing, China (No. Z221100007422121), Natural Science Foundation of China (Nos. 32141005, 32000530, and 32371035), Beijing Nova Program (No. 20230484246), the National Key R&D Program of China (No. 2022YFA1105502), Young Elite Scientists Sponsorship Program by CAST (No. 2023QNRC001, YESS20240790), the national key R&D projects (No. 32141005), Beijing Natural Science Foundation (Nos. L223021, L232122, 7244305, L222133, and 7252018), Capital’s Funds for Health Improvement and Research (No. CFH 2024-1-5021), and Noncommunicable Chronic Diseases-National Science and Technology Major Project (No. 2026ZD0557100, 2026ZD0557105).