Left ventricular hypertrophy detection using electrocardiographic signal

Area of Science:

• Cardiology
• Biomedical Engineering
• Artificial Intelligence in Medicine

Background:

• Left Ventricular Hypertrophy (LVH) signifies subclinical organ damage and is linked to cardiovascular disease incidence.
• Electrocardiogram (ECG) is a cost-effective, non-invasive tool for preliminary heart disease diagnosis.
• Current ECG criteria for LVH rely on voltage thresholds of RS peaks.

Purpose of the Study:

• To develop and evaluate an automated system for detecting LVH using ECG signals.
• To compare the performance of the developed system against traditional ECG criteria and existing AI models.

Main Methods:

• A two-step system was developed: 1) Extraction of 24 features from R-peak and S-valley amplitudes of 12-lead ECG, followed by ECG beat segmentation. 2) Training a back propagation neural network (BPN) with these features.
• Echocardiography (ECHO) served as the gold standard for LVH diagnosis.
• The BPN model was trained and tested on a dataset from a Taiwanese population, including 173 LVH cases and 1466 segmented ECG cycles.

Main Results:

• The BPN model achieved high testing performance: accuracy (0.961), precision (0.958), sensitivity (0.966), and specificity (0.956).
• The developed BPN system demonstrated superior detection performance compared to 7 traditional ECG criteria methods.
• Performance also surpassed previously reported ECG-based artificial intelligence (AI) models for LVH detection.

Conclusions:

• The developed BPN system offers a highly accurate and efficient method for detecting LVH from ECG signals.
• This AI-driven approach shows significant potential for improving early diagnosis of cardiovascular conditions.
• The system's performance suggests a valuable advancement over conventional diagnostic methods for LVH.