Non-invasive kidney diagnosis: A multimodal transformer system for diabetic nephropathy diagnosis via retinal imaging

Abstract It remains a formidable challenge to differentiate between diabetic nephropathy (DN) and non-diabetic renal disease (NDRD) without resorting to kidney biopsy, resulting in missed opportunities for targeted interventions that may significantly alter the prognosis of NDRD. 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 the superior eneralizability 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). On the other hand, the accuracy of antidiastole detection by nephrologists can be improved by 21%, through visualization assistance of the DL system. Therefore, this paper lays a foundation for automatically differentiating DN and NDRD without biopsy.