Evaluation of Classification Tree Application for Estimation of ECG Signal Quality

Abstract

Smart textile used to monitor ECG is very comfortable but requires severe quality assessment during data analysis. The aim of this research was to develop algorithm able to accept or reject ECG episodes based on its quality using simple and easy to compute signal quality indices (SQIs) for single channel ECG analysis device. Classification trees using CART algorithm were used to test ten SQIs and build final optimized tree. Optimized tree revealed best performing parameters including moving differentiating filter, root mean square and standard deviation quality indices. Whole classification tree performance was substantially better compared with algorithm using only one SQI.