An inherently explainable AI trained on 1,015 expert-annotated prostate tissue images achieved strong Gleason pattern segmentation while providing interpretable outputs and addressing interobserver variability in pathology.