AI ECG Model Interpretability and Confounding Bias Assessment

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

AI-based ECG analysis models lack interpretability and are prone to confounding bias, limiting their acceptance and usability in clinical environments due to the unpredictable utilization of features and criteria.

Innovation Solution

A method is introduced to extract interpretable criteria from an AI-based ECG analysis model, allowing for the assessment of bias, thereby enhancing model interpretability and usability by outputting these criteria and bias assessments to user devices.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Productivity

If AI-based ECG analysis models are used, then diagnostic accuracy and automation are improved, but interpretability and reliability deteriorate due to unpredictable feature utilization and confounding bias

Engineering Contradiction:

Improvediagnostic automationVSAvoidmodel interpretability

Core Design Contradiction:

ProductivityVSReliability

Solution Approach 1:

The patent introduces an intermediary system that sits between the AI model and the clinician. This intermediary extracts and presents the top K most influential features and criteria that the AI model used to generate its diagnostic output, making the black-box AI decision-making process transparent and interpretable for clinical use

Inventive Principle:

Principle #24Intermediary (Mediator)

Solution Approach 2:

The system implements feedback by continuously monitoring and assessing confounding bias in the AI model's feature utilization. It identifies when the model relies on inappropriate or spurious features and provides feedback to adjust or retrain the model, ensuring reliable and interpretable diagnostic reasoning

Inventive Principle:

Principle #23Feedback

2Productivity

If AI-based ECG analysis models are deployed, then efficiency is improved, but bias and fairness deteriorate due to confounding factors in model training

Engineering Contradiction:

Improveanalysis efficiencyVSAvoidconfounding bias

Core Design Contradiction:

ProductivityVSObject-affected harmful factors

Solution Approach 1:

The patent performs preliminary bias assessment before deploying the AI model in clinical settings. It pre-identifies potential confounding factors and evaluates the model's susceptibility to bias, allowing for preventive measures to be taken before the model causes harmful decisions

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system converts the harmful effect of confounding bias into a beneficial diagnostic tool by using the identified biased features as warnings to clinicians. When the model shows reliance on spurious features, the system alerts clinicians to interpret the output with caution, turning the bias detection capability into a safety mechanism

Inventive Principle:

Principle #22Blessing in disguise (Convert harm into benefit)

Data Source

PatentUS20260058012A1Systems and methods for adding interpretability to and assessing bias of an ECG analysis model

Publication Date: 2026.02.26 GE PRECISION HEALTHCARE LLC

AI summary

Systems and methods for adding interpretability to and assessing bias of an ECG analysis model are herein provided. In one example, a method comprises: obtaining a diagnostic output from an AI-based ECG analysis model on an ECG dataset; extracting interpretable criteria from the ECG dataset for a target of the diagnostic output of the AI-based ECG analysis model to predict an output for the ECG dataset based on the extracted criteria; determining one or more characteristics of the extracted criteria; assessing the output from the ECG analysis model for bias based on a comparison between the output of the ECG analysis model and the predicted output; and outputting the one or more characteristics and the bias assessment to a user device.