Vectorcardiography-based software for revealing non-evident features from ECG signals

MALDONADO KEVIN; BARCOS JAVIER; AVACA HORACIO; SPAGNUOLO DAMIAN; SCHIAVONE MIGUEL; CÁCERES MONIÉ CESAR; FERNÁNDEZ JUAN; PLUCHE NATALÍ; TELLO SANTA CRUZ IVÁN; CHARA OSVALDO

Congreso; 16th World Congress of Arrhythmias; 2019

Background: Since its development, vectorcardiography took a secondary place behindelectrocardiography regarding to electrical activity evaluation in the clinical practice. With increasingcomputing capability and availability, there is a resurgence of vectorcardiography techniques althoughclinical applications are not well established yet.Objective: To develop a cardiac electrical activity vectorcardiography-based analysis method andsoftware.Methods: 12 leads ECGs were digitally recorded at 1000 Hz. Data of interest were selected manually orautomatically depending on analysis type. After preprocessing, vectorcardiograms were synthesizedfrom ECGs using a transformation matrix and the following variables were obtained: signal area, meanvoltage, mean instantaneous vector speed, curvature, duration, QRS octant and intra and inter-patientsignal correlation. For automatic ECG database analysis, we studied 90 healthy volunteers (49 fromPhysionet?s PTB database and 41 obtained by us) and 200 randomly picked ECGs from patients whounderwent routine evaluation for suspected or known chronic hypertension.Results: The software adequately identified electrical signals from multiple ECGs in databases andperformed vectorcardiogram reconstruction and analysis. Key features were obtained automatically,allowing fast and precise characterization of large amounts of data. In control individuals, a linearrelation was found between mean spatial vector magnitude and mean instantaneous vector speed(slope=76.16 s -1 , intercept=9.24 mV/ms, R 2 =0.81). The slope of this relation was defined as the controlpopulation time constant (s -1 ). Control intra-patient ventricular depolarization signal correlation(Pearson?s) was 0.9985 (ranging from 0.9904 to 0.9997, showing the low variability between eachpatient ventricular depolarization signals. When hypertensive patients database was analyzed this linearrelation was maintained, although dispersion of data increased significantly probably due to thepresence of comorbidities affecting ventricular conduction (slope=38.80 s -1 , intercept=22.84 mV/ms,R 2 =0.23).Conclusions: Digital ECG processing and vectorcardiogram reconstruction allowed us to assess non-evident signal features, which may be of clinical relevance. This approach could be a promising tool inclinical cardiology applications such as massive ECG database analysis, electrical synchrony evaluationand arrhythmia ablation.