Self-supervised Normality Learning and Divergence Vector-guided Model Merging for Zero-shot Congenital Heart Disease Detection in Fetal Ultrasound Videos

2025-03-10

AI Summary

Zero-shot detection of congenital heart disease (CHD) in fetal ultrasound videos poses significant challenges due to data scarcity and strict privacy constraints across clinical centers. Method: We propose a privacy-preserving, data-free federated framework wherein each center trains solely on local normal fetal echocardiographic videos. A sparse Tube-based self-supervised learning paradigm is introduced to extract healthy spatiotemporal representations. CHD detection is reformulated as “normality learning,” and a novel DivMerge strategy is proposed—leveraging divergence-guided vector aggregation to enable federated model fusion without sharing raw data or labels. The method integrates spatiotemporal feature disentanglement, self-distillation loss, and unsupervised anomaly scoring. Results: Evaluated on real-world data from five hospitals, the fused model achieves 23.77% higher accuracy and 30.13% higher F1-score on external test cohorts compared to single-center baselines, demonstrating strong zero-shot generalization while guaranteeing strict data privacy.

Technology Category

• Machine Learning: Distributed Machine Learning & Federated Learning
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Fact-Checking / Misinformation Detection (NLP Focus)

Application Category

• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Systems and Infrastructure for Web, Mobile and WoT: Decentralized Web and Fediverse systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Congenital Heart Disease (CHD) is one of the leading causes of fetal mortality, yet the scarcity of labeled CHD data and strict privacy regulations surrounding fetal ultrasound (US) imaging present significant challenges for the development of deep learning-based models for CHD detection. Centralised collection of large real-world datasets for rare conditions, such as CHD, from large populations requires significant co-ordination and resource. In addition, data governance rules increasingly prevent data sharing between sites. To address these challenges, we introduce, for the first time, a novel privacy-preserving, zero-shot CHD detection framework that formulates CHD detection as a normality modeling problem integrated with model merging. In our framework dubbed Sparse Tube Ultrasound Distillation (STUD), each hospital site first trains a sparse video tube-based self-supervised video anomaly detection (VAD) model on normal fetal heart US clips with self-distillation loss. This enables site-specific models to independently learn the distribution of healthy cases. To aggregate knowledge across the decentralized models while maintaining privacy, we propose a Divergence Vector-Guided Model Merging approach, DivMerge, that combines site-specific models into a single VAD model without data exchange. Our approach preserves domain-agnostic rich spatio-temporal representations, ensuring generalization to unseen CHD cases. We evaluated our approach on real-world fetal US data collected from 5 hospital sites. Our merged model outperformed site-specific models by 23.77% and 30.13% in accuracy and F1-score respectively on external test sets.

Problem

• Detecting congenital heart disease in fetal ultrasound videos.
• Overcoming data scarcity and privacy in CHD detection.
• Merging decentralized models without sharing data.

Innovation

• Self-supervised normality learning for anomaly detection
• Divergence Vector-Guided Model Merging (DivMerge)
• Privacy-preserving zero-shot CHD detection framework