Title :
Discriminating between supraventricular and ventricular tachycardias from EGM onset analysis
Author :
Rojo-Álvarez, José L. ; Arenal-Maíz, Ángel ; Artes- Rodriguez, A.
Author_Institution :
Escuela Politenica Superior, Univ. Carlos III, Madrid, Spain
Abstract :
We hypothesize that the analysis of the ventricular electrogram onset (EGM onset) can discriminate between SVT and VT to obtain a simultaneous increase in sensitivity and specificity. We discuss our analysis of EGMs obtained during SVT and VT together with their preceding SRs in 38 SVT and 68 VT far field records from 16 patients. The, algorithmic implementation and the preprocessing tasks were performed through the support vector method (SVM), avoiding the overfitting by means of the statistical bootstrap resampling. To improve the safety for an individual patient, two new methods of incremental learning, based on the SVM, are proposed and tested on an independent set of spontaneous arrhythmia episodes.
Keywords :
defibrillators; electrocardiography; learning (artificial intelligence); learning automata; medical expert systems; medical signal processing; pacemakers; radial basis function networks; signal classification; signal sampling; time-frequency analysis; RBF classifier; algorithmic implementation; arrhythmia treatment; automatic tasks; classification problem; cost learning; heuristic featuring; implantable cardioverter defibrillators; incremental learning; margin learning; optimum time interval; pacing capability; sensitivity; specificity; spontaneous arrhythmia episodes; statistical bootstrap resampling; support vector method; supraventricular tachycardia; time-frequency analysis; ventricular depolarization; ventricular electrogram onset; ventricular tachycardia; Batteries; Cardiology; Computerized monitoring; Electric shock; Heart rate; Medical treatment; Rhythm; Sensitivity and specificity; Strontium; Support vector machines; Algorithms; Artificial Intelligence; Bundle of His; Diagnosis, Differential; Electrocardiography; Evaluation Studies as Topic; Feedback; Humans; Models, Cardiovascular; Models, Statistical; Pattern Recognition, Automated; Sample Size; Sensitivity and Specificity; Signal Processing, Computer-Assisted; Tachycardia, Supraventricular; Tachycardia, Ventricular;
Journal_Title :
Engineering in Medicine and Biology Magazine, IEEE
