Noninvasive Diagnosis of Coronary Artery Disease Based on Instantaneous Frequency of Diastolic Murmurs and SVM

Noninvasive diagnosis method of coronary artery disease is proposed based on the instantaneous frequency estimate of diastolic murmurs and support vector machine(SVM) classifier. Hilbert-Huang transform is studied to analyze the instantaneous frequency of diastolic murmurs. The weighted instantaneous frequency is introduced. Using statistical quantities of weighted instantaneous frequency forms feature vectors. In end the SVM classifier is build for diagnosis of CAD. 77 clinical recordings (40 abnormal and 37 normal) are categorized as two groups: training sets (20 abnormal and 20 normal) which are employed to train the classifier and test sets (20 abnormal and 17 normal) which are employed to test the classifier. 17 of 20 abnormal and all normal are correctly diagnosed. The results show that the method offers a promising prospect for early noninvasive diagnosis of CAD