Synergistic Foundation Models for Semi-Supervised Fetal Cardiac Ultrasound Analysis: SAM-Med2D Boundary Refinement and DINOv3 Semantic Enhancement

2026-05-19

AI Summary

This study addresses the challenge of joint segmentation and classification in fetal echocardiography under severe label scarcity by proposing a semi-supervised multi-task learning framework. Built upon the EchoCare backbone, the method integrates SAM-Med2D for boundary refinement and leverages DINOv3 to enhance pseudo-label quality. A novel view-specific hard masking mechanism and a two-stage optimization strategy are introduced: the first stage employs exponential moving average (EMA) to boost segmentation performance, while the second stage freezes segmentation parameters and resets the classification head to restore discriminative capability. Evaluated on the FETUS 2026 benchmark, the model achieves a Dice coefficient of 79.99%, a normalized surface distance of 61.62%, and an F1 score of 41.20%, significantly outperforming existing approaches.

Abstract

We present a semi-supervised framework for joint segmentation and classification of fetal cardiac ultrasound images. Built upon the EchoCare multi-task backbone, our method integrates SAM-Med2D for boundary refinement and leverages DINOv3 to enhance pseudo-label quality. We introduce view-specific hard masking along with a two-stage optimization strategy: an EMA phase to consolidate segmentation capabilities, followed by a Classification Fine-Tuning phase that freezes segmentation parameters and resets the classification head to recover classification performance without compromising segmentation gains. Evaluated on the FETUS 2026 leaderboard, our method achieves a Dice Similarity Coefficient at 79.99%, Normalized Surface Distance at 61.62%, and F1-score at 41.20%, validating the effectiveness of our approach for prenatal congenital heart disease screening. Source code is publicly available at: https://github.com/2826056177/zcst_fetus2026.

Problem

• fetal cardiac ultrasound
• semi-supervised learning
• image segmentation
• image classification
• congenital heart disease screening

Innovation

• SAM-Med2D
• DINOv3
• semi-supervised learning
• boundary refinement
• two-stage optimization