Federated Neural Network Training for Echocardiogram Analysis

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Current automated systems for analyzing echocardiogram images are limited in recognizing and analyzing both 2D and Doppler modality images, unable to distinguish between similar heart diseases, and require manual intervention for diagnosis and prognosis, leading to inefficiencies and incomplete data utilization.

Innovation Solution

A software-based clinical workflow that uses neural networks to automatically recognize and analyze both 2D and Doppler modality echocardiographic images, enabling automated measurements, diagnosis, and prognosis, with a federated training platform for distributed training and validation across multiple patient cohorts.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If manual annotation and measurement of cardiac structures is performed by clinicians, then diagnostic accuracy can be maintained through expert interpretation, but the process becomes highly time-consuming and labor-intensive, resulting in over 95% of available images never being annotated or quantified

Engineering Contradiction:

Improvediagnostic accuracyVSAvoidimage processing throughput

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system enables automated self-service processing of echocardiogram images through neural networks that automatically annotate cardiac structures, measure dimensions, and generate diagnostic reports without requiring manual clinician intervention for each image, thereby processing the overwhelming majority of stored DICOMs that previously remained unannotated

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent replaces the manual mechanical process of clinician annotation and measurement with an automated computational system using neural networks and image processing algorithms, substituting human labor with machine-based automated analysis while maintaining diagnostic quality

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Productivity

If automated systems are implemented to increase processing efficiency, then productivity improves, but the systems are limited to recognizing only 2D images and cannot distinguish between similar-looking heart diseases, requiring manual intervention

Engineering Contradiction:

Improveimage processing throughputVSAvoiddisease differentiation capability

Core Design Contradiction:

ProductivityVSAdaptability or versatility

Solution Approach 1:

The automated system is designed with multi-functionality to perform multiple tasks: recognizing both 2D and Doppler modality images, differentiating between various heart diseases including HFpEF and HFREF, extracting clinical measurements, and generating comprehensive diagnostic reports, thereby eliminating the need for manual intervention while maintaining versatility

Inventive Principle:

Principle #6Universality (Multi-functionality)

Solution Approach 2:

The system uses parameter changes in neural network architecture and training approaches to enhance disease differentiation capability, incorporating multiple imaging modalities and analyzing various physiological parameters to distinguish between similar-looking heart diseases that previous automated systems could not differentiate

Inventive Principle:

Principle #35Parameter changes

3Measurement precision

If comprehensive analysis of both 2D and Doppler modality images is performed, then diagnostic accuracy and disease differentiation improve, but the complexity of the analysis system and training requirements increase significantly

Engineering Contradiction:

Improvedisease diagnosis accuracyVSAvoidsystem and training complexity

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The analysis system is segmented into specialized neural network components: one branch processes 2D images for anatomical structure analysis, another branch processes Doppler modality images for functional assessment, and additional modules handle disease classification and measurement extraction, allowing each component to be optimized independently while maintaining overall diagnostic accuracy

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS11301996B2Training neural networks of an automatic clinical workflow that recognizes and analyzes 2D and doppler modality echocardiogram images

Publication Date: 2022.04.12 EKO AI PTE LTD

AI summary

A method for training neural networks of an automated workflow performed by a software component executing on a server in communication with remote computers at respective laboratories includes downloading and installing a client and a set of neural networks to a first remote computer of a first laboratory, the client accessing the echocardiogram image files of the first laboratory to train the set of neural networks and to upload a first trained set of neural networks to the server. The process continues until the client and the second trained set of neural networks is downloaded and installed to a last remote computer of a last laboratory, the client accessing the echocardiogram image files of the last laboratory to continue to train the second trained set of neural networks and to upload a final trained set of neural networks to the server.