Support Vector Black-box Interpretation in Ventricular Arrhythmia Discrimination

Abstract

In this article we propose two SVM-oriented analyses and their use in building two new differential diagnosis algorithms based on the ventricular EGM onset criterion. The following approaches are suggested: 1) a geometrical analysis of the input feature space and its relationship to the critical samples (i.e., the support vectors); 2) a study of the relevance of the activation time state. As was demonstrated in the companion article, an incremental learning procedure should be used for each algorithmic implementation in order to reduce the inter-patient variability as new information about the patient (i.e., new arrhythmia episodes) becomes available. Note that the records in BaseC(training control group) and Base D (independent test group) have been described in the companion article.