A Decision-making Method for Defibrillation Based on SVM and Opt-AMSA

Since the existing methods for defibrillation decision have yet to be perfected, we attempt to redesign a decision-making method for detecting ventricular fibrillation (VF) and deciding the timing of defibrillation in this paper. Our method is combined with Support Vector Machine (SVM) and Opt-AMSA, aiming at rhythm classification and optimal decision on defibrillation timing respectively. It mainly includes two parts, rhythm classification and decision on defibrillation timing. We firstly convert the electrocardiograph (ECG) signal from time domain to frequency domain by Fourier transform and select a serious of magnitudes of frequency components in a certain frequency range as feature vectors, which are ready for rhythm classification. Then, Support Vector Machine (SVM) is applied to classify the ECG signals in frequency domain so as to recognize VF after feature selection. Once the VF is recognized, Opt-AMSA will make the optimal decision on defibrillation timing. In our experiment we involved an ECG dataset containing 148 samples and got an ideal decision accuracy, which ultimately proved that our decision-making method based on SVM and Opt-ASMA performs more effectively than others.