Echocardiogram Classification with Machine Learning for EROA

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current echocardiogram analysis for valvular heart diseases relies heavily on manual expertise, is time-consuming, error-prone, and varies widely between operators, while existing automated solutions are limited by using 2-D cross-sectional images and lack comprehensive lesion representation.

Innovation Solution

A computer-implemented method using machine learning algorithms to analyze echocardiogram images, including classification and segmentation of different modalities and structures, to automatically determine metrics like Effective Regurgitant Orifice Area (EROA) and Regurgitant Volume, utilizing neural networks for modality classification, structure identification, and landmark detection.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual analysis by experienced cardiologists is used, then diagnostic accuracy is improved, but time consumption and operational complexity increase significantly

Engineering Contradiction:
Improvediagnostic accuracyVSAvoidtime consumption
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary automated analysis of echocardiogram images using machine learning models to identify potential abnormalities and prepare preliminary diagnoses before cardiologist review, reducing the time required for manual analysis while maintaining diagnostic accuracy

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

An automated image analysis system with machine learning algorithms serves as an intermediary between the echocardiogram images and the cardiologist, pre-processing and highlighting key features to accelerate the diagnostic process without replacing the expert clinician

Inventive Principle:
Principle #24Intermediary (Mediator)

2Reliability

If manual analysis by multiple operators is used, then comprehensive assessment is achieved, but variability in interpretations and error rates increase

Engineering Contradiction:
Improveinterpretation consistencyVSAvoidoperational complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The system performs self-service automated analysis of echocardiogram images using machine learning models that consistently identify and measure cardiac structures, eliminating operator variability and reducing errors while maintaining comprehensive assessment capabilities

Inventive Principle:
Principle #25Self-service

Solution Approach 2:

The system transforms subjective manual measurements into objective quantitative parameters through automated image analysis, standardizing the assessment process across different cases and operators while reducing interpretation variability

Inventive Principle:
Principle #35Parameter changes

3Productivity

If 2-D cross-sectional images are used for automated analysis, then processing speed is improved, but diagnostic accuracy and lesion representation deteriorate

Engineering Contradiction:
Improveprocessing speedVSAvoidlesion representation accuracy
Core Design Contradiction:
ProductivityVSMeasurement precision

Solution Approach 1:

The system transitions from 2-D cross-sectional image analysis to 3-D volumetric echocardiogram data analysis, enabling more accurate lesion representation and comprehensive cardiac structure assessment while maintaining automated processing capabilities

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

Solution Approach 2:

The system integrates multiple imaging modalities and data types (3-D volumetric data, 2-D images, and associated metadata) into a composite analysis framework, combining the advantages of different data representations to improve both processing efficiency and diagnostic accuracy

Inventive Principle:
Principle #40Composite materials

4Productivity

If fully automated analysis is implemented, then time efficiency and consistency are improved, but system complexity and development difficulty increase

Engineering Contradiction:
Improveanalysis efficiencyVSAvoidsystem complexity
Core Design Contradiction:
ProductivityVSDevice complexity

Solution Approach 1:

The automated analysis system is segmented into modular machine learning models that perform specific tasks (image classification, structure identification, measurement, and report generation) independently, reducing overall system complexity while maintaining high analysis efficiency

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The system employs universal machine learning models trained on diverse echocardiogram data that can perform multiple diagnostic functions across different cardiac conditions and imaging protocols, reducing the need for condition-specific specialized systems

Inventive Principle:
Principle #6Universality (Multi-functionality)

Data Source

PatentEP4608278B1Echocardiogram classification with machine learning
Publication Date: 2025.11.19 LIGENCE UAB
  • EP4608278B1 patent drawingFigure 1
  • EP4608278B1 patent drawingFigure 2
  • EP4608278B1 patent drawingFigure 3

AI summary

A method for analysing a structure within a patient's heart comprises the steps of receiving a patient study comprising a plurality of transthoracic echocardiography and/or transesophageal echocardiography images of a patient's heart obtained by an ultrasound device. The method comprises extracting respective sets of pixel data from the images, and using trained machine learning algorithms to analyse the sets of pixel data to a) identify images captured using a colour Doppler modality and assign one or more of these images to said structure, b) identify images captured using spectral Doppler modality and assign one or more of these images to said structure, c) determine a flow convergence zone radius R from an analysis of the colour Doppler modality images assigned to the structure; d) determine a maximum gradient of structure, PVreg, from an analysis of the spectral Doppler modality images assigned to the structure. The flow convergence zone radius, R, the maximum gradient of structure PVreg, and a Nyquist value Va identified for the colour Doppler modality, are combined to determine an Effective Regurgitant Orifice Area, EROA for the structure.