Optimizing ECG to detect echocardiographic left ventricular hypertrophy with computer-based ECG data and machine...

Area of Science:

• Cardiology
• Medical Informatics
• Machine Learning

Background:

• Left ventricular hypertrophy (LVH) detected by echocardiography (Echo-LVH) is a significant predictor of mortality.
• Electrocardiogram (ECG) analysis, particularly with advanced algorithms like Philips DXL-16, offers insights into cardiac electrical activity.
• Machine learning (ML) techniques present opportunities for developing novel diagnostic criteria for Echo-LVH.

Purpose of the Study:

• To identify a novel combination of ECG parameters for predicting Echo-LVH.
• To develop and validate a new computer-based model, the Cardiac Hypertrophy Computer-based model (CHCM), for Echo-LVH detection.

Main Methods:

• Extracted 458 ECG parameters from the Philips DXL-16 algorithm in patients with and without Echo-LVH.
• Utilized the C5.0 ML algorithm for training, testing, and validating the CHCM.
• Compared the diagnostic performance of the CHCM against established state-of-the-art criteria.

Main Results:

• The CHCM incorporates T voltage in lead I, peak-to-peak QRS duration in aVL, and peak-to-peak QRS duration in aVF.
• In the primary cohort, the CHCM achieved an accuracy of 70.5% (sensitivity 74.3%, specificity 68.7%).
• External validation demonstrated a CHCM accuracy of 63.5% (sensitivity 42%, specificity 82.9%), outperforming several established criteria.

Conclusions:

• A novel set of ECG parameters, identified through ML analysis of computer-based ECG data, can predict Echo-LVH.
• The CHCM effectively classifies patients based on Echo-LVH with repolarization abnormalities or increased voltage.
• The CHCM demonstrates comparable accuracy and slightly improved sensitivity over current state-of-the-art criteria for Echo-LVH detection.