Streamlining the Histopathologic Workflow in Diabetic Kidney Disease with Artificial Intelligence

In this study, we accelerated and enhanced pathology readouts in a real-life pharmaceutical industry setting. We show that AI-assisted scoring reduced pathologists' workload and expedited study assessments. Self-supervised learning captured intrinsic properties of kidney morphology without expert annotation and reduced expert bias and translational discrepancies, greatly facilitating translational activities in drug development for patients with diabetic kidney disease.