Neural Network for Lead-Invariant ECG Analysis

Solution Overview

Problem

Conventional ECG systems require multiple leads for accurate readings, which limits their use in wearable and portable devices that often utilize fewer leads or a single lead.

Innovation Solution

A machine learning system utilizing a neural network architecture that includes a feature extraction sub-neural network and a feature fusing sub-neural network to process ECG data from any number of leads, allowing for lead-invariant analysis and prediction of cardiac conditions.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional ECG systems use multiple leads for accurate readings, then measurement precision is improved, but device complexity and adaptability deteriorate

Engineering Contradiction:

ImproveECG reading accuracyVSAvoidcompatibility with wearable devices

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system changes the parameter of lead configuration from fixed 12-lead to variable lead counts (1-12 leads). The neural network is trained to process ECG inputs with any number of leads, transforming the system from requiring a specific fixed configuration to being adaptable to different lead configurations used in wearable devices.

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The neural network model achieves universality by being able to process ECG data from any number of leads (1 to 12). This single model serves multiple functions: it works with single-lead wearables, multi-lead portable devices, and conventional 12-lead ECG systems, eliminating the need for separate models for different lead configurations.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Device complexity

If single lead or few leads are used in wearable devices, then device complexity is reduced, but measurement precision deteriorates

Engineering Contradiction:

Improvenumber of electrodesVSAvoidECG reading accuracy

Core Design Contradiction:

Device complexityVSMeasurement precision

Solution Approach 1:

The patent replaces the mechanical/electrical approach of using multiple electrodes with a computational approach using neural networks. Instead of relying on having multiple physical leads to achieve accuracy, the system uses a neural network that can process signals from fewer leads, substituting physical complexity with computational intelligence.

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system changes the parameter of lead configuration from fixed 12-lead to variable lead counts (1-12 leads). The neural network is trained to process ECG inputs with any number of leads, transforming the system from requiring a specific fixed configuration to being adaptable to different lead configurations used in wearable devices.

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If lead-specific models are used for processing ECG data, then measurement precision is improved, but adaptability deteriorates

Engineering Contradiction:

Improvedisease detection accuracyVSAvoidlead-invariant processing capability

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The neural network model achieves universality by being able to process ECG data from any number of leads (1 to 12). This single model serves multiple functions: it works with single-lead wearables, multi-lead portable devices, and conventional 12-lead ECG systems, eliminating the need for separate models for different lead configurations.

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

Instead of creating separate specialized models for each lead configuration and then trying to make them work together, the invention inverts the approach by creating a single universal model that inherently handles all lead configurations. This inversion of the modeling strategy achieves both precision and adaptability simultaneously.

Inventive Principle:

Principle #13The other way round (Inversion)

Data Source

PatentUS20250069759A1Machine-learning for processing lead-invariant electrocardiogram inputs

Publication Date: 2025.02.27 MAYO FOUNDATION FOR MEDICAL EDUCATION & RESEARCH

AI summary

Provided herein are methods, systems, and computer program products for the detection and evaluation of cardiac condition in a lead-invariant manner.