QRS complex analysis using wavelet transform and two layered Self-Organizing Map

Many kinds of methods have been developed to classify QRS complex in Holter electrocardiogram. However, the accuracy of these methods dose not fully satisfy the clinical needs. In this paper, we developed an automated classification methods using a wavelet transform and two-layered Self-Organizing Map (SOM) to improve the accuracy. A discrete wavelet coefficient is used as a characteristic parameter for the heart beat and the two-layered SOM is used for classification. First, each beat is divided in eight sections and the discrete wavelet coefficients of level 1–5 are calculated using a Haar mother wavelet for each section. By learning these characteristics in the first SOM, each section is classified automatically. Second, the QRS complexes are reconstructed as a line of the classified class in the first-SOM and classified by the second SOM. We evaluated our method using MIT-BIH Arrhythmia database of 16 cases (32,032 beats) and compared it with the accuracy of a standard cross correlation coefficient method. The classification error rate of the correlation coefficient method and proposed method is 0.82% and 0.39% respectively. We confirmed that the accuracy of our method for the QRS complex analysis has significantly improved.