Detection and Classification of Cardiac Arrhythmias.

摘要

Expansion of indications for Implantable Cardioverter Defibrillators has led to a significant increase in the number of patients receiving ICDs and the number of lives saved due to ICD therapy. However, inappropriate shocks due to misclassification of supraventricular and ventricular arrhythmias are still a major problem among current ICDs.;The problem is due to the fact that supraventricular arrhythmias - which do not require shock therapy - share some of the characteristics of ventricular arrhythmias - which require shock therapy - and can lead to unnecessary therapy delivery. These inappropriate shocks are the most common adverse event among ICD recipients. Documented rates of inappropriate therapy range from 11% to 41% and may result in a significant decrease in quality of life in patients with ICDs. Worse, inappropriate therapy can be proarrhythmic, making the optimization of detection algorithms critical. Because of hardware restrictions these algorithms should have low computational complexity and power consumption, which makes developing them more difficult.;The goal of this dissertation is employing signal processing, data analysis and pattern recognition techniques to develop and test new ICD rhythm discrimination algorithms that can potentially improve the classification rate of current ICD devices and help decrease the number of inappropriate shocks for ICD patients. New rhythm discrimination algorithms based on Dynamic Time Warping, estimated covariance matrices and Proximal Support Vector Machines are proposed. All three proposed algorithms are applicable on both single and dual chamber ICDs and meet the computational complexity restrictions of ICD devices.