BanffNET, a Deep Learning System for Comprehensive Histological Lesion Quantification in Kidney Transplant Biopsies

Abstract Accurate, reproducible interpretation of kidney allograft biopsies is critical for the diagnosis of graft injury and for informing prognosis and clinical management. The international Banff classification is a consensus diagnostic system based on semiquantitative histological lesion scoring according to either lesion extent or severity in kidney transplant biopsies. However, pathologist scoring is limited by interobserver variability, constrained scalability, and the inherent nature of the scoring system itself. Here we present BanffNET, a weakly supervised, probabilistic deep learning framework that combines self-supervised feature extraction with a novel Bayesian multiple-instance learning framework to predict (continuously) the full spectrum of Banff lesion scores directly from whole-slide images (WSIs). Using lesion-specific aggregation functions tailored to localized (modeling severity) and diffuse histological lesions (modeling extent), BanffNET generates interpretable, patch-level probability maps and calibrated slide-level scores. BanffNET’s performance was assessed relative to consensus, biological correlates of rejection and clinical outcome, demonstrating superior consistency, transportability and generalization. Trained on 7,533 WSIs from three cohorts, BanffNET demonstrates consistent performance on 12,687 WSIs across five external validation cohorts, matching or surpassing individual expert pathologists across lesion assessments. BanffNET scores align more closely than pathologist Banff scores with molecular profiles of rejection, offering an objective, transparent, biologically grounded framework for computational pathology with relevance beyond kidney transplantation.