Diagnostic and prognostic value of ECG-predicted hypertension-mediated left ventricular hypertrophy using machine...

Area of Science:

• Cardiology
• Medical Informatics
• Biomedical Engineering

Background:

• Hypertension-induced left ventricular hypertrophy (LVH) presents diverse phenotypes with distinct prognoses.
• Cardiac magnetic resonance (CMR) identifies these LVH phenotypes but is not widely accessible.
• Electrocardiogram (ECG) is a routine diagnostic tool, yet its ability to differentiate LVH phenotypes remains unclear.

Purpose of the Study:

• To develop and validate machine learning (ML) models for classifying hypertension-mediated LVH phenotypes using ECG data.
• To assess the association of ECG-predicted LVH phenotypes with incident cardiovascular outcomes.

Main Methods:

• Utilized ECG biomarkers from 20,439 hypertensive patients in the UK Biobank (UKB).
• Developed classification models (logistic regression, SVM, random forest) integrating ECG and clinical data.
• Externally validated models in the Study of Health in Pomerania (SHIP) cohort and assessed associations with major adverse cardiovascular events (MACE) and heart failure.

Main Results:

• Machine learning models demonstrated comparable classification performance, with SVM achieving an accuracy of 0.79.
• The models showed superior prediction for eccentric LVH.
• ECG-predicted eccentric LVH was significantly associated with an increased risk of heart failure in the UKB test set.

Conclusions:

• ECG-based ML classifiers offer a promising, accessible screening strategy for early detection of hypertension-mediated LVH phenotypes.
• This approach can aid in identifying patients at higher risk for adverse cardiovascular events.