Echocardiographic Diagnosis of Hypertrophic Cardiomyopathy by Machine Learning

Area of Science:

• Cardiology
• Medical Imaging
• Machine Learning

Background:

• Hypertrophic cardiomyopathy (HCM) is a significant cardiovascular condition.
• Accurate and early detection of HCM is crucial for patient management.
• Echocardiography is a primary imaging modality for cardiac assessment.

Purpose of the Study:

• To develop and evaluate machine learning (ML) models for automated HCM identification.
• To assess the performance of ML models using standard echocardiographic metrics.
• To determine the added value of strain measurements in ML-based HCM detection.

Main Methods:

• Development of four random forest ML models using a large case-control cohort (5548 HCM patients, 16,973 controls).
• Models were trained using demographic data and 16 standard echocardiographic metrics, with variations in strain data inclusion.
• Ten-fold cross-validation was employed for model training and validation.

Main Results:

• All four ML models demonstrated high performance in identifying HCM cases.
• Models incorporating strain data (global, averaged, or individual) achieved superior area under the curve (AUC) values (0.96-0.98) compared to the model without strain data (AUC 0.92).
• Model 4, utilizing 17 individual strains, showed the highest performance (AUC 0.98).

Conclusions:

• Machine learning tools effectively identify HCM from echocardiographic metrics.
• The inclusion of strain data in ML models significantly enhances the performance of HCM detection.
• ML-based analysis of echocardiographic data, particularly with strain parameters, offers a promising approach for automated HCM recognition.