Two layered classification using qualitative and quantitative attributes for QRS complex analysis

QRS complex classification in Holter electrocardiogram have been developed using the correlation coefficient methods. However, the accuracy of this traditional classification is not fully satisfied the clinical needs. In this paper, we propose a two-layered classification using qualitative and quantitative attributes. In the first layer, 24 components in a FFT power spectrum for each beat are calculated as the quantitative attributes and are classified using K-means algorithm. In the second layer, the numbers of low, middle and high peaks before/after an R wave are computed as the qualitative attributes and are also classified by the same way. We evaluated our method for ten cases from MIT-BIH arrhythmia databases and compared with a standard cross correlation coefficient method. The classification error rate of the correlation coefficient method and proposed method is 1.10% and 0.79%. We confirmed that the accuracy in our method for the QRS complex analysis is significantly improved.