EVIL: Evidential Inference Learning for Trustworthy Semi-supervised Medical Image Segmentation

TL;DR

This paper introduces EVIL, a novel evidential inference framework using Dempster-Shafer Theory for semi-supervised medical image segmentation, providing accurate uncertainty quantification and trustworthy pseudo labels with theoretical guarantees.

Contribution

EVIL integrates Dempster-Shafer Theory into semi-supervised segmentation, enabling reliable uncertainty estimation and improved generalization with minimal labeled data.

Findings

Achieves competitive performance on public datasets.

Provides theoretically guaranteed uncertainty quantification.

Enhances segmentation accuracy with pseudo labels.

Abstract

Recently, uncertainty-aware methods have attracted increasing attention in semi-supervised medical image segmentation. However, current methods usually suffer from the drawback that it is difficult to balance the computational cost, estimation accuracy, and theoretical support in a unified framework. To alleviate this problem, we introduce the Dempster-Shafer Theory of Evidence (DST) into semi-supervised medical image segmentation, dubbed Evidential Inference Learning (EVIL). EVIL provides a theoretically guaranteed solution to infer accurate uncertainty quantification in a single forward pass. Trustworthy pseudo labels on unlabeled data are generated after uncertainty estimation. The recently proposed consistency regularization-based training paradigm is adopted in our framework, which enforces the consistency on the perturbed predictions to enhance the generalization with few labeled data. Experimental results show that EVIL achieves competitive performance in comparison with several state-of-the-art methods on the public dataset.