System for the recognition and diagnostics of cardiac arrhythmias applied to the identification of wide complex tachycardia from the ECG

Abstract

A software system as a support for the differential diagnostics of wide complex tachycardia based on the Bayesian diagnostic clinical methodology is presented. The system has different modules to: a) Detect complexes, with Sensibility (Sn) of 94% and Predictive Positive value (VP+) of 97.5%. b) Compute the duration of the complex, with root mean square error (ERMS) of 25.2 ms. c) Determinate the electric axis of the QRS complex, whit RMS error of 5.89º. d) Determinate the branch block morphology, with Sn 89.9% and VP+ 93.3% and d) Classify the QRS morphologies, where complete classification was obtained. The used technique was based in the zero crossings, and singular values (maxima and minima) in the Wavelet Transform. The classification system was developed using artificial neural networks.