ECG Attention Mapping for Explainable Atrial Fibrillation Detection

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Deep learning models for atrial fibrillation detection in ECG signals lack interpretability, making it difficult for clinical experts to understand the decision-making process, which hinders their adoption in practical clinical scenarios.

Innovation Solution

A method and system that converts 1D ECG signals to 2D images, applies a deep learning classifier model with clinical domain knowledge, generates class activation maps, and calculates attention scores to provide clinical expert-like explanations for the classification decisions, aligning with how cardiologists analyze ECG signals.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If deep learning models are used for autonomous identification of atrial fibrillation in ECG, then detection accuracy is improved, but interpretability deteriorates making the models opaque black boxes

Engineering Contradiction:
Improvedetection accuracyVSAvoidinterpretability
Core Design Contradiction:
Measurement precisionVSLoss of information

Solution Approach 1:

The patent introduces class activation maps (CAM) as an intermediary mechanism that bridges the gap between the deep learning model's internal decision-making process and clinical interpretability. The CAM visualizes which regions of the ECG signal contribute most to the model's classification decision, providing explainable insights without compromising the model's accurate detection capabilities.

Inventive Principle:
Principle #24Intermediary (Mediator)

Solution Approach 2:

The patent transforms the one-dimensional ECG signal into a two-dimensional representation to enable the application of attribution methods and generate class activation maps. This dimensional transformation allows for visual interpretation of model decisions while maintaining the temporal characteristics of the original signal, thus resolving the contradiction between accuracy and interpretability.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Productivity

If deep learning models provide autonomous classification without explanation, then productivity is improved, but clinical acceptance deteriorates due to lack of explanation for decision-making

Engineering Contradiction:
Improveautonomous classification efficiencyVSAvoidclinical acceptance
Core Design Contradiction:
ProductivityVSEase of operation

Solution Approach 1:

The patent implements a feedback mechanism where class activation maps provide explanatory information back to clinicians about why the model made a particular classification. This feedback loop maintains autonomous classification efficiency while enhancing clinical acceptance by making the model's reasoning transparent and interpretable for medical professionals.

Inventive Principle:
Principle #23Feedback

Data Source

PatentEP4678104A1Method and system for explaining decision-making process of model detecting atrial fibrillation in electrocardiogram waves
Publication Date: 2026.01.14 TATA CONSULTANCY SERVICES LTD
  • EP4678104A1 patent drawingFigure 1~2
  • EP4678104A1 patent drawingFigure 3
  • EP4678104A1 patent drawingFigure 4

AI summary

Current approaches for atrial fibrillation (AF) detection use deep learning models which remain opaque. In particular, they lack in providing explanation of why this particular decision (around existence of AF) has been made, thereby making it unacceptable to clinical domain experts. Present disclosure provides method and system for explaining decision-making process of deep learning models used for detecting AF in ECG waves. The system receives ECG signal which is converted into two-dimensional (2D) representation which further helps in classification of diagnosis condition from ECG signal using classifier model. Thereafter, system generates class activation maps (CAM) to find attention scores and finally uses these attention scores, to identify top R-R intervals where classifier model is placing greater emphasis. Further, system converts ECG image into ECG signal which also converts CAM into attention wave. Finally, system uses ECG signal and attention wave to generate clinical expert like explanations for class label prediction.