A Method based on Wavelet Packet Decomposition and Memory Antibody Clone Hybrid Clustering for Diagnosis of Clinical Heart Disorders

摘要

In this study, a method based on wavelet packet (WP) decomposition and memory antibody clone hybrid clustering (MACHC) for diagnosis of clinical heart disorder (HD) is proposed. From considering the fact that the frequency ranges of the normal and heart diseases sound are different from each other, the wavelet packet decomposition (WPD) at level 4 is used to split the frequency bandwidths of the clinical heart sound (CHS) signals. And then the WP energy (WPE) with the frequency information through the range of CHS signals is calculated. WPEs at the terminal nodes from (4,0) to (4,11) are selected and two parameters maxWPE and minWPE are used as the features. Furthermore, the MACHC technique is employed as the identification tool to classify the normal sound and CHSs. Finally, the performances of this method are evaluated in 156 samples that contain 42 normal and 114 abnormal subjects for clinical HDs. The results show that this technique is useful and effective to detect CHSs. The validation of the proposed method is measured by using the sensitivity, specificity and accuracy parameters. Over 94.91% sensitivity, 100% specificity and 96.25% accuracy rate are obtained.