다중 클래스 SVM을 이용한 부정맥 신호 분류

Abstract

Electrocardiogram(ECG) Analysis and arrhythmia recognition are critical for diagnosis and treatment of cardiac disorders. Analysis and experiments of ECG are very important in the diagnosis of heart disease. Arrhythmias represent a serious threat to the patient recovering from acute myocardial infarction and myocardial ischemia. In particular, ventricular tachycardia(VT) and ventricular fibrillation(VF) are life threatening arrhythmias with sudden cardiac death. There is a need for quick detection of arrhythmias. Other arrhythmias like atrial premature contraction(APC), premature ventricular contraction(PVC) and superventricular tachycardia(SVT) are not as fatal as VF, but are important in diagnosing the disorders of the heart. The reliable detection of these arrhythmias constitutes a challenge for a cardiovascular diagnostic system. Consequently, various techniques have been studied to classify arrhythmias which mostly classify two or three arrhythmias. A technique is proposed to classify normal sinus rhythm(NSR) and APC, PVC, SVT, VT and VF. This paper present emprical mode decomposition(EMD) based ECG analysis and classification of cardiac arrhythmia using multi-class support vector machine(SVM). EMD method as a preprocessor and a direct ECG feature extraction has been employed. The main objective of EMD was to decompose the signal into a set of function, defined by the signal itself, named Intrinsic Mode Functions(IMFs), which preserve the inherent properties of the original signal. Since the decomposition is based on the local time scale of the signal, it is not only applicable to nonlinear and non-stationary processes but also useful in biomedical signals like electrocardiogram. AR modeling is another feature extraction method which reduce the number of classification parameters into a reasonably small set for a meaningful classification. AR coefficients extracted from Burg's algorithm have been used for arrhythmia classification in conjunction with other features. The AR order of four was sufficient for modeling the ECG signals. Multi-class SVMs are usually implemented by combining several two-class SVMs. The one-against-all method using winner-takes-all strategy and the one-against-one method implemented by max-wins voting are popularly used for this purpose. In this paper, one-against-one method is hired. The results show that multi-class SVM is useful for classification of cardiac arrhythmias. Parameters for SVM classifker, and , were chosen as 200 and 1 respectively by trial and error experiments. Each average performance of detecting NSR, APC, PVC, SVT, VT and VF was 98.11%, 97.39%, 96.42%, 98.27%, 96.82% and 99.41%. When the performance of SVM classifier was compared to that of Backpropagation, it showed similar or high values. Consequently, we could find that the proposed input features and SVM classifier would one of the most useful algorithm for cardiac arrhythmia classification. Further validation of the proposed method will yield acceptable results for clinical implementation.