Exploring Bayesian Deep Learning Uncertainty Measures for Segmentation of New Lesions in Longitudinal MRIs

Nazanin Mohammadi Sepahvand, Raghav Mehta, Douglas Lorne Arnold, Doina Precup, Tal Arbel

Keywords: Multiple Sclerosis, New and enlarging lesions, longitudinal MRI, Bayesian Deep Learning.

Track: short paper

Paper Type: well-validated application

Abstract:

In this paper, we develop a modified U-Net architecture to accurately segment new and
enlarging lesions in longitudinal MRI, based on multi-modal MRI inputs, as well as subtrac-
tion images between timepoints, in the context of large-scale clinical trial data for patients
with Multiple Sclerosis (MS). We explore whether MC-Dropout measures of uncertainty
lead to confident assertions when the network output is correct, and are uncertain when
incorrect, thereby permitting their integration into clinical workflows and downstream in-
ference tasks.