Disease Progression of Hypertrophic Cardiomyopathy: Modeling Using Machine Learning.

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Cardiovascular disorders cause 30% of global deaths.
• Hypertrophic cardiomyopathy (HCM) affects 1 in 500 young adults, posing a risk for sudden cardiac death (SCD).
• Current methods lack long-term clinical status prediction for HCM patients.

Purpose of the Study:

• Develop a novel machine learning (ML) tool for predicting HCM disease progression.
• Forecast adverse cardiac remodeling over a 10-year period.
• Enhance patient management through long-term clinical status prediction.

Main Methods:

• Utilized 6 predictive regression models for key clinical characteristics.
• Independently predicted left atrial size, left atrial volume, ejection fraction, NYHA class, and ventricular diameters.
• Employed Shapley additive explanation for model interpretability.

Main Results:

• ML models demonstrated lower predictive error than human experts (average 0.34 vs. experts, 0.22 vs. consortium).
• Semisupervised learning and virtual patient data improved predictive accuracy.
• The best random forest model achieved an R² increase from 0.3 to 0.6.

Conclusions:

• ML models showed favorable performance compared to experts for 5 of 6 predicted clinical targets.
• Expert validation confirmed the models' clinical utility.
• The tool offers a promising approach for long-term HCM management.