Machine learning algorithms for predicting arrhythmic events in Hypertrophic Cardiomyopathy: limited enhancement...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Hypertrophic Cardiomyopathy (HCM) is a genetic heart condition associated with increased risk of sudden cardiac death.
• Current risk stratification models for HCM, like the ESC risk score, have limitations in accurately predicting arrhythmic events.
• Late gadolinium enhancement (LGE) on cardiac magnetic resonance (CMR) is a known imaging biomarker for risk stratification in HCM.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model using common clinical features to predict arrhythmic events in HCM patients.
• To compare the performance of the developed ML model against the established ESC HCM risk score and LGE quantification.

Main Methods:

• A post-hoc analysis was conducted on data from 531 international HCM patients who underwent CMR.
• Clinical, echocardiographic, and CMR variables, including LGE quantification, were used to train and test various ML models.
• The Random Forest (RF) model was selected as the best performing model and its predictive performance was assessed using AUC.

Main Results:

• The Random Forest (RF) ML model demonstrated strong performance in predicting arrhythmic events (AUC = 0.78), significantly outperforming the ESC HCM risk score (AUC = 0.64).
• The ML model's predictive performance was comparable to LGE quantification alone (AUC = 0.76), with no significant incremental improvement.
• The study identified a composite endpoint of sudden cardiac death, aborted SCD, and sustained ventricular tachycardia.

Conclusions:

• Machine learning models integrating clinical variables can significantly improve the prediction of arrhythmic events in HCM compared to current clinical risk scores.
• The incremental value of clinical ML models over LGE imaging alone is limited, highlighting the strong prognostic power of LGE.
• These findings are exploratory and suggest further investigation into integrating ML with imaging biomarkers for HCM risk stratification.