npj Digital Medicine (Jan 2025)
A multimodal transformer system for noninvasive diabetic nephropathy diagnosis via retinal imaging
- Zheyi Dong,
- Xiaofei Wang,
- Sai Pan,
- Taohan Weng,
- Xiaoniao Chen,
- Shuangshuang Jiang,
- Ying Li,
- Zonghua Wang,
- Xueying Cao,
- Qian Wang,
- Pu Chen,
- Lai Jiang,
- Guangyan Cai,
- Li Zhang,
- Yong Wang,
- Jinkui Yang,
- Yani He,
- Hongli Lin,
- Jie Wu,
- Li Tang,
- Jianhui Zhou,
- Shengxi Li,
- Zhaohui Li,
- Yibing Fu,
- Xinyue Yu,
- Yanqiu Geng,
- Yingjie Zhang,
- Liqiang Wang,
- Mai Xu,
- Xiangmei Chen
Affiliations
- Zheyi Dong
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Xiaofei Wang
- Department of Clinical Neurosciences, University of Cambridge
- Sai Pan
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Taohan Weng
- School of Electronic and Information Engineering, Beihang University
- Xiaoniao Chen
- Senior Department of Ophthalmology, Third Medical Center of Chinese PLA General Hospital
- Shuangshuang Jiang
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Ying Li
- Department of Ophthalmology, First Medical Center of Chinese PLA General Hospital
- Zonghua Wang
- Department of Ophthalmology, Seventh Medical Center of Chinese PLA General Hospital
- Xueying Cao
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Qian Wang
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Pu Chen
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Lai Jiang
- School of Electronic and Information Engineering, Beihang University
- Guangyan Cai
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Li Zhang
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Yong Wang
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Jinkui Yang
- Beijing Key Laboratory of Diabetes Research and Care, Beijing Diabetes Institute, Beijing Tongren Hospital, Capital Medical University
- Yani He
- Department of nephrology, Daping Hospital, Army Medical University
- Hongli Lin
- The First Affiliated Hospital of Dalian Medical UniversityKey Laboratory of Kidney Disease, Center for the Transformation Medicine of Kidney Disease
- Jie Wu
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Li Tang
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Jianhui Zhou
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Shengxi Li
- School of Electronic and Information Engineering, Beihang University
- Zhaohui Li
- School of Clinical Medicine, Guangdong Pharmaceutical University
- Yibing Fu
- School of Electronic and Information Engineering, Beihang University
- Xinyue Yu
- School of Clinical Medicine, Guangdong Pharmaceutical University
- Yanqiu Geng
- Department of Nephrology, Third Medical Center of Chinese PLA General Hospital
- Yingjie Zhang
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- Liqiang Wang
- Senior Department of Ophthalmology, Third Medical Center of Chinese PLA General Hospital
- Mai Xu
- School of Electronic and Information Engineering, Beihang University
- Xiangmei Chen
- Department of Nephrology, First Medical Center of Chinese People’s Liberation Army General Hospital
- DOI
- https://doi.org/10.1038/s41746-024-01393-1
- Journal volume & issue
-
Vol. 8,
no. 1
pp. 1 – 14
Abstract
Abstract Differentiating between diabetic nephropathy (DN) and non-diabetic renal disease (NDRD) without a kidney biopsy remains a major challenge, often leading to missed opportunities for targeted treatments that could greatly improve NDRD outcomes. To reform the traditional biopsy-all diagnostic paradigm and avoid unnecessary biopsy, we developed a transformer-based deep learning (DL) system for detecting DN and NDRD upon non-invasive multi-modal data of fundus images and clinical characteristics. Our Trans-MUF achieved an AUC of 0.980 (95% CI: 0.979 to 0.980) over the internal retrospective set and also had superior generalizability over a prospective dataset (AUC: 0.989, 95% CI: 0.987 to 0.990) and a multicenter, cross-machine and multi-operator dataset (AUC: 0.932, 95% CI: 0.931 to 0.939). Moreover, the nephrologists‘ diagnosis accuracy can be improved by 21%, through visualization assistance of the DL system. This paper lays a foundation for automatically differentiating DN and NDRD without biopsy. (Registry name: Correlation Study Between Clinical Phenotype and Pathology of Type 2 Diabetic Nephropathy. ID: NCT03865914. Date: 2017-11-30).
