Multi-Lead ECG Region-of-Interest Analysis for Quantifiable AI Explanations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing deep learning models for cardiovascular disease classification using multi-lead electrocardiograms provide results without explaining the reasoning behind their decisions, leading to a lack of trust from medical professionals and the need for qualitative explanations.
Innovation Solution
A processor-implemented method and system that generate quantifiable explanations for multi-lead electrocardiograms by using a trained classifier and explainer model to compute contribution values for each data point, filter and normalize these values, identify prominent leads, and analyze their contribution to disease classes, enabling quantifiable and qualitative explanations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning models are used for cardiovascular disease classification, then classification accuracy is improved, but explainability deteriorates
Solution Approach 1:
The patent segments the ECG signal into 12 separate leads and further divides each lead into multiple time windows. The explanation system then analyzes each segment independently to identify which specific leads and time periods contribute most to the disease classification, providing granular explainability while maintaining overall accuracy.
Solution Approach 2:
The patent introduces an intermediary explanation generation system that acts as a mediator between the deep learning classifier and the user. This system takes the classification results and translates them into human-readable explanations by identifying and highlighting the specific ECG leads and time windows that were most influential in the classification decision.
2Ease of operation
If qualitative explanations are provided through visual representations, then ease of understanding is improved, but measurement precision deteriorates
Solution Approach 1:
The patent transforms qualitative visual explanations into quantitative parameter-based explanations by calculating and reporting specific contribution values for each lead and time window. These numerical parameters precisely measure the degree to which each ECG component influenced the classification, providing both accuracy and interpretability.
Solution Approach 2:
The patent replaces purely visual explanation mechanisms with a hybrid system that incorporates numerical contribution values. Instead of relying solely on visual highlighting, the system provides quantifiable metrics that precisely measure and communicate the importance of each ECG lead and time period, enhancing both precision and understandability.
Data Source
AI summary
Deep Learning (DL) performs well in cardiovascular disease (CVD) classification using 12-lead Electrocardiogram (ECG). However, explainable artificial intelligence in CVD classification still remains largely qualitative. Embodiments of the present disclosure provide systems and methods that implement a Region of Interest (ROI) based quantifiable explanation for multi-lead ECG. CVD specific post-processing steps are added, to increase the explanation performance. Furthermore, the system enables selection of an optimal DL model, within the performance space defined by classification, explanation, and time-complexity.