Uncertainty-Aware Retinal Vessel Segmentation via Ensemble Distillation

Abstract

We propose Ensemble Distillation as an approach to uncertainty estimation in retinal vessel segmentation, distilling the knowledge of multiple ensemble models into a single model. Evaluated on the DRIVE and FIVES datasets, this method achieves comparable performance via calibration and segmentation metrics while significantly reducing computational complexity, offering an efficient and reliable approach for uncertainty estimation in medical imaging applications.

Jeremiah Fadugba (Jerofad)

Data Scientist | Machine Learning Engineer

My current research interests include Medical Imaging, Machine Learning, Deep Learning, and Trustworthy ML.