SS-CADA: A Semi-Supervised Cross-Anatomy Domain Adaptation for Coronary Artery Segmentation

TL;DR

This paper introduces SS-CADA, a semi-supervised method that leverages limited annotations and cross-domain data to improve coronary artery segmentation in X-ray angiograms, addressing cross-anatomy domain shift.

Contribution

The paper proposes a novel semi-supervised domain adaptation framework with vesselness-specific batch normalization and self-ensembling to reduce annotation needs in coronary artery segmentation.

Findings

Achieves accurate coronary artery segmentation with limited labeled data.

Effectively addresses cross-anatomy domain shift in medical imaging.

Outperforms existing methods in segmentation accuracy.

Abstract

The segmentation of coronary arteries by convolutional neural network is promising yet requires a large amount of labor-intensive manual annotations. Transferring knowledge from retinal vessels in widely-available public labeled fundus images (FIs) has a potential to reduce the annotation requirement for coronary artery segmentation in X-ray angiograms (XAs) due to their common tubular structures. However, it is challenged by the cross-anatomy domain shift due to the intrinsically different vesselness characteristics in different anatomical regions under even different imaging protocols. To solve this problem, we propose a Semi-Supervised Cross-Anatomy Domain Adaptation (SS-CADA) which requires only limited annotations for coronary arteries in XAs. With the supervision from a small number of labeled XAs and publicly available labeled FIs, we propose a vesselness-specific batch normalization (VSBN) to individually normalize feature maps for them considering their different cross-anatomic vesselness characteristics. In addition, to further facilitate the annotation efficiency, we employ a self-ensembling mean-teacher (SEMT) to exploit abundant unlabeled XAs by imposing a prediction consistency constraint. Extensive experiments show that our SS-CADA is able to solve the challenging cross-anatomy domain shift, achieving accurate segmentation for coronary arteries given only a small number of labeled XAs.