Abstract: We developed a novel approach to address label noise and low inter-annotator consensus in medical image classification. Applying a Bayesian neural network (BNN), we classified Meibomian ...