Hierarchical Consistency Regularized Mean Teacher for Semi-supervised 3D Left Atrium Segmentation

TL;DR

This paper proposes a hierarchical consistency regularized mean teacher framework for semi-supervised 3D left atrium segmentation, effectively leveraging limited annotations to achieve performance comparable to fully supervised methods.

Contribution

It introduces a novel hierarchical consistency regularization approach within a mean teacher framework for semi-supervised 3D medical image segmentation.

Findings

Achieves competitive performance with full annotation methods

Outperforms existing semi-supervised segmentation techniques

Demonstrates effectiveness of hierarchical consistency regularization

Abstract

Deep learning has achieved promising segmentation performance on 3D left atrium MR images. However, annotations for segmentation tasks are expensive, costly and difficult to obtain. In this paper, we introduce a novel hierarchical consistency regularized mean teacher framework for 3D left atrium segmentation. In each iteration, the student model is optimized by multi-scale deep supervision and hierarchical consistency regularization, concurrently. Extensive experiments have shown that our method achieves competitive performance as compared with full annotation, outperforming other state-of-the-art semi-supervised segmentation methods.