Identifying Premature Ventricular Complexes from Outflow Tracts Based on PVC Configuration: A Machine Learning...

Area of Science:

• Cardiology
• Medical Data Analytics
• Signal Processing

Background:

• Traditional methods for determining the site of origin (SOO) of premature ventricular complexes (PVCs) from surface electrocardiograms (ECGs) lack advanced data analysis.
• Visual inspection of ECGs may miss subtle signals crucial for accurate SOO determination.

Purpose of the Study:

• To apply advanced data analytics to characterize PVCs and determine their SOO.
• To leverage machine learning for improved accuracy in PVC origin identification.

Main Methods:

• Unsupervised machine learning cluster analysis was performed on 12-lead ECGs from 338 individuals.
• Specific ECG indexes derived from lead V1, V2, and V3 were analyzed.
• Results were compared against a composite criterion for SOO.

Main Results:

• Specific criteria using V1S and V2S values effectively identified left ventricular outflow tract (LVOT) origin (sensitivity 95.4%, specificity 97.5%).
• Right bundle branch block (RBBB) and left bundle branch block (LBBB) configurations, combined with specific amplitude criteria in leads V1-V3, indicated LVOT or right ventricular outflow tract (RVOT) origins.
• Distinct ECG patterns reliably differentiated between LVOT and RVOT origins.

Conclusions:

• Novel data analytic techniques offer a non-invasive approach for identifying PVC SOO.
• These advanced methods enhance the clinician's ability to interpret 12-lead ECGs for PVC origin.
• The findings support the integration of machine learning in routine cardiac diagnostics.