TransECG: Leveraging Transformers for Explainable ECG Re-identification Risk Analysis

2025-03-11

AI Summary

Cross-institutional ECG signal sharing poses significant privacy risks due to the potential for individual re-identification. While existing deep learning re-identification models achieve high accuracy, they lack interpretability, hindering identification of the underlying physiological features driving recognition. To address this, we propose the first interpretable framework for ECG re-identification, innovatively adapting Vision Transformers (ViTs) for ECG time-series modeling and leveraging attention-map visualization to quantitatively assess the contributions of key physiological components—including the R-wave, QRS complex, and P-R interval—to gender, age, and individual ID classification. Evaluated on four real-world multicenter datasets comprising 87 subjects, our model achieves 89.9% accuracy for gender, 89.9% for age, and 88.6% for identity re-identification. Crucially, interpretability analysis reveals the R-wave (58.29%) and P-R interval (46.29%) as the most discriminative regions—demonstrating simultaneous advances in both performance and physiological interpretability.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Privacy
• Planning, Routing, and Scheduling: Activity and Plan Recognition

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Security and Privacy: Data transparency and provenance
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Electrocardiogram (ECG) signals are widely shared across multiple clinical applications for diagnosis, health monitoring, and biometric authentication. While valuable for healthcare, they also carry unique biometric identifiers that pose privacy risks, especially when ECG data shared across multiple entities. These risks are amplified in shared environments, where re-identification threats can compromise patient privacy. Existing deep learning re-identification models prioritize accuracy but lack explainability, making it challenging to understand how the unique biometric characteristics encoded within ECG signals are recognized and utilized for identification. Without these insights, despite high accuracy, developing secure and trustable ECG data-sharing frameworks remains difficult, especially in diverse, multi-source environments. In this work, we introduce TransECG, a Vision Transformer (ViT)-based method that uses attention mechanisms to pinpoint critical ECG segments associated with re-identification tasks like gender, age, and participant ID. Our approach demonstrates high accuracy (89.9% for gender, 89.9% for age, and 88.6% for ID re-identification) across four real-world datasets with 87 participants. Importantly, we provide key insights into ECG components such as the R-wave, QRS complex, and P-Q interval in re-identification. For example, in the gender classification, the R wave contributed 58.29% to the model's attention, while in the age classification, the P-R interval contributed 46.29%. By combining high predictive performance with enhanced explainability, TransECG provides a robust solution for privacy-conscious ECG data sharing, supporting the development of secure and trusted healthcare data environment.

Problem

• Analyzing ECG signals for privacy risks in shared environments.
• Developing explainable deep learning models for ECG re-identification.
• Identifying critical ECG segments linked to biometric identifiers.

Innovation

• Vision Transformer for ECG analysis
• Attention mechanisms identify critical ECG segments
• High accuracy with explainable biometric insights