AnomExpert: Identifying and Selecting Anatomical Planes for Prenatal Ultrasound Anomaly Diagnosis

2026-07-14

AI Summary

This work proposes AnomExpert, a weakly supervised framework that jointly addresses anatomical plane identification and disease-relevant plane selection in multiplanar prenatal ultrasound images using only case-level labels. Without requiring any plane-level annotations, AnomExpert introduces learnable anatomical plane prototypes to structurally represent unordered image instances and incorporates a pathology-aware sparse attention mechanism to automatically identify the most diagnostically informative planes for specific congenital anomalies. Built upon a ViT-small backbone and prototype-driven multiple instance learning, the method achieves 86.9% accuracy and 84.2% F1 score on a multicenter dataset of 3,654 cases, significantly outperforming nine state-of-the-art multiple instance learning approaches while maintaining parameter efficiency.

Abstract

Life-limiting congenital anomalies require accurate prenatal diagnosis for appropriate clinical decision-making. Prenatal ultrasound (US) examinations involve multiple anatomical planes, and diagnosis depends on identifying anatomical planes and selecting diagnostically relevant planes for each anomaly. Existing automated methods either rely on plane-level annotations or aggregate heterogeneous images without explicitly modeling these diagnostic capabilities. We propose AnomExpert, a prototype-driven framework for prenatal US anomaly diagnosis using only case-level supervision. AnomExpert introduces learnable plane prototypes to organize unordered images into latent representations corresponding to anatomical planes without requiring plane annotations. A disease-aware sparse selection mechanism further selects diagnostically relevant planes for each anomaly. Experiments on a multi-center dataset of 3,654 cases show that AnomExpert consistently outperforms nine representative multi-instance learning methods. Using a ViT-small backbone, it achieves 86.9% accuracy and 84.2% F1-score while maintaining parameter efficiency. These findings indicate that modeling anatomical plane identification and disease-specific plane selection improves weakly supervised multi-plane prenatal US anomaly classification. The code is available at https://github.com/TIanCat/AnomExpert.

Problem

• prenatal ultrasound
• anatomical planes
• congenital anomalies
• weakly supervised learning
• multi-instance learning

Innovation

• learnable plane prototypes
• disease-aware sparse selection
• weakly supervised learning
• anatomical plane identification
• multi-plane ultrasound diagnosis