Reduction of Ventricular Tachycardia False Alarms Using Multivariate Statistical Process Control and Frequency Analysis

Abstract

In this paper, we present a false alarm reduction algorithm for Ventricular Tachycardia (VT) arrhythmias in intensive care unit (ICU) using multivariate statistical process control (MSPC) and frequency analysis of electrocardiogram (ECG) signal. First, we decompose the ECG signal into three different frequency bands. The ECG beats are detected, and VT beats are labeled. In the next step, several features consist of time domain features, bispectrum features, and Poincaré plot features are extracted from ECG Signal The extracted feature vector of each ECG beat is monitored using MSPC for detecting anomalies. The performance of the proposed method is evaluated using the Ventricular Tachycardia cases of 2015 Physionet challenge database. This dataset consists of 2 ECG channel, arterial blood pressure (ABP) and/or photoplethysmograph (PPG) signal, and an alarm annotation for each record. The obtained sensitivity and specificity were 86.5%, and 80.7% respectively. We have also investigated the advantage of using ABP signal in improving the results of false alarm reduction. For this purpose, some biological features are extracted from ABP and used as an extra feature vector. Results show that using ABP signal can improve the performance of the algorithm.