Artificial intelligence (AI) and machine learning (ML) technologies are increasingly being investigated for their potential to transform diagnostic and prognostic capabilities in nephrology. Applications span image analysis of renal biopsies, predictive modeling of CKD progression, AKI risk stratification, and dialysis outcome optimization. A comprehensive review published in Nature Reviews Nephrology (Komorowski & Celi, 2017) outlined the theoretical and practical frameworks for AI deployment in kidney disease management.
ISN (International Society of Nephrology) has recognized AI integration as a priority for future nephrology practice. A landmark study by TomaĊĦev et al. (DeepMind, Nature, 2019) described a deep learning model trained on US Veterans Affairs electronic health records that predicted AKI onset up to 48 hours before clinical diagnosis with an AUC of 0.92, substantially outperforming conventional risk scoring tools in a hospital population of over 700,000 patients.
In renal pathology, deep learning algorithms applied to digitized kidney biopsy specimens have shown promise for automated classification of glomerular lesions. Zheng et al. (KidneyNet, Journal of the American Society of Nephrology, 2021) demonstrated that a convolutional neural network achieved pathologist-level accuracy in classifying IgA nephropathy Oxford MEST-C scores, potentially reducing inter-observer variability in multicenter studies.
Predictive models for CKD progression incorporating clinical variables, biomarker trajectories, and genomic data have been developed and externally validated. The Kidney Failure Risk Equation (KFRE), published by Tangri et al. in JAMA (2011) and validated in over 30 countries, uses four to eight clinical variables to predict two- and five-year risk of ESRD with high discrimination, and has been endorsed by the International Society of Nephrology as a practical risk stratification tool in clinical guidelines.
ISN has emphasized that AI tools must be developed with attention to racial and ethnic representation in training datasets. A commentary in Journal of the American Society of Nephrology (Bowe et al., 2021) documented how race-based eGFR equations embedded in prior AI training pipelines perpetuated health disparities. The transition to race-free CKD-EPI 2021 equations, supported by ISN, reflects the society’s broader commitment to algorithmic equity in kidney care.