Machine Learning Assessment of Left Ventricular Diastolic Function Based on Electrocardiographic Features.

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Left ventricular (LV) diastolic dysfunction is a key factor in heart failure pathophysiology.
• Current clinical tools for early detection of diastolic dysfunction lack precision before echocardiography.

Purpose of the Study:

• To develop machine-learning (ML) models for quantitative estimation of myocardial relaxation.
• To utilize clinical and electrocardiography (ECG) variables for early detection of LV diastolic dysfunction.

Main Methods:

• A multicenter prospective study involved 1,202 subjects across 4 North American institutions.
• ML models were trained and tested on signal-processed ECG, traditional ECG, and clinical features.
• Model generalizability was assessed using an external test set from a separate institution.

Main Results:

• The ML model accurately predicted LV relaxation velocities (e') in both internal and external test sets (mean absolute error: 1.46 and 1.93 cm/s).
• The model demonstrated strong discrimination for abnormal myocardial relaxation and LV diastolic/systolic dysfunction (AUCs ranging from 0.75 to 0.84).
• Estimated e' predicted LV diastolic dysfunction with high accuracy (AUCs of 0.88 and 0.94).

Conclusions:

• Quantitative prediction of myocardial relaxation is feasible using accessible clinical and ECG data.
• This cost-effective method can serve as an initial clinical step for assessing LV dysfunction.
• The approach may aid in the early diagnosis and management of heart failure.