Automatic Supporting System for Regionalization of Ventricular Tachycardia Exit Site in Implantable Defibrillators

• DOI: 10.1371/journal.pone.0124514
• Publication Date: 2015-04-24T14:07:59Z
• Authors: Margarita Sanromá...; Inmaculada Mora-J...; Jesús Almendral; Arcadio García-Al...; José Luis Rojo-Ál...
• Keywords: Male; Artificial intelligence; Support Vector Machine; Support vector machine; Cardiac Resynchronization Therapy in Heart Failure; Science; Cardiology; Catheter Ablation of Cardiac Arrhythmias; Pattern recognition (psychology); Electrocardiography; 03 medical and health sciences; 0302 clinical medicine; Implantable Cardioverter-Defibrillator; Health Sciences; Machine learning; Humans; Aged; Left Ventricular Dysfunction; Q; R; Reproducibility of Results; Ventricular tachycardia; Middle Aged; Molecular Mechanisms of Cardiac Arrhythmias; Computer science; Defibrillators, Implantable; Ventricular Tachycardia; Echocardiography; Tachycardia, Ventricular; Medicine; Female; Cardiac Electrophysiology; Cardiology and Cardiovascular Medicine; Research Article

Electrograms stored in Implantable Cardioverter Defibrillators (ICD-EGM) have been proven to convey useful information for roughly determining the anatomical location of Left Ventricular Tachycardia exit site (LVTES). Our aim here was evaluate possibilities from a machine learning system intended provide an estimation LVTES region with use ICD-EGM situation where 12-lead electrocardiogram ventricular tachycardia are not available. Several techniques were specifically designed and benchmarked, both classification (such as Neural Networks (NN), Support Vector Machines (SVM)) regression (Kernel Ridge Regression) problem statements. Classifiers evaluated by using accuracy rates identification controlled number regions, approach quality studied terms spatial resolution. We analyzed 23 patients (18±10 EGM per patient) during left pacing simultaneous recording coordinates electrode navigation system. feature sets extracted (consisting times voltages) shown more discriminative than raw waveform. Among classifiers, SVM performed slightly better NN. In accordance previous clinical works, average resolution about 3 cm, our system, which allows it support faster determination ablation procedures. The proposed also provides framework suitable driving design improved performance future systems.