Explaining deep learning for ECG using time-localized clusters

2025-09-18

AI Summary

Deep learning achieves strong performance in ECG analysis, yet its “black-box” nature severely hinders clinical trustworthiness and mechanistic knowledge discovery. To address this, we propose Temporal Localized Clustering (TLC), the first method to partition and cluster internal CNN representations along the time axis while integrating uncertainty quantification, enabling interpretable, segment-wise ECG analysis. TLC requires no additional annotations; it localizes contributions of physiologically meaningful regions—such as P, QRS, and T waves—to model predictions and visualizes both decision rationale and confidence levels. Experiments demonstrate that TLC not only recovers established electrophysiological patterns but also identifies novel, prognostically relevant localized representation clusters associated with heart failure. This work establishes the first ECG deep learning interpretability framework that jointly supports structured explanation, uncertainty-aware modeling, and clinically actionable interpretation—significantly enhancing transparency and practical utility of AI-assisted diagnosis.

Technology Category

• Machine Learning: Calibration & Uncertainty Quantification
• Computer Vision: Interpretability, Explainability, and Transparency
• Reasoning under Uncertainty: Uncertainty Representations

Application Category

• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Web query analysis, representation and understanding

Abstract

Deep learning has significantly advanced electrocardiogram (ECG) analysis, enabling automatic annotation, disease screening, and prognosis beyond traditional clinical capabilities. However, understanding these models remains a challenge, limiting interpretation and gaining knowledge from these developments. In this work, we propose a novel interpretability method for convolutional neural networks applied to ECG analysis. Our approach extracts time-localized clusters from the model's internal representations, segmenting the ECG according to the learned characteristics while quantifying the uncertainty of these representations. This allows us to visualize how different waveform regions contribute to the model's predictions and assess the certainty of its decisions. By providing a structured and interpretable view of deep learning models for ECG, our method enhances trust in AI-driven diagnostics and facilitates the discovery of clinically relevant electrophysiological patterns.

Problem

• Interpreting deep learning models for ECG analysis
• Explaining model predictions using time-localized clusters
• Quantifying uncertainty in ECG waveform representations

Innovation

• Time-localized clusters extraction method
• Uncertainty quantification for ECG representations
• Visualization of waveform contributions to predictions