The role of MRI physics in brain segmentation CNNs: achieving acquisition invariance and instructive uncertainties

TL;DR

This paper introduces a physics-informed, uncertainty-aware CNN for brain segmentation that achieves robustness to different MRI acquisition parameters and generalizes well to unseen data, with reliable uncertainty estimates.

Contribution

It presents a novel MRI physics-based augmentation and feature stratification approach that enhances acquisition invariance and uncertainty quantification in brain segmentation CNNs.

Findings

Improved robustness to site and sequence variability.

Accurate extrapolation to out-of-distribution samples.

Enhanced uncertainty calibration for volumetric estimates.

Abstract

Being able to adequately process and combine data arising from different sites is crucial in neuroimaging, but is difficult, owing to site, sequence and acquisition-parameter dependent biases. It is important therefore to design algorithms that are not only robust to images of differing contrasts, but also be able to generalise well to unseen ones, with a quantifiable measure of uncertainty. In this paper we demonstrate the efficacy of a physics-informed, uncertainty-aware, segmentation network that employs augmentation-time MR simulations and homogeneous batch feature stratification to achieve acquisition invariance. We show that the proposed approach also accurately extrapolates to out-of-distribution sequence samples, providing well calibrated volumetric bounds on these. We demonstrate a significant improvement in terms of coefficients of variation, backed by uncertainty based volumetric validation.