Shockable Arrhythmia Detection Using Sequential Minimal Optimization, and Support Vector Machine

Ventricular arrhythmias (VAs) are potentially fatal arrhythmias and the primary cause of sudden cardiac arrest (SCA). To prevent SCA, immediate treatment is defibrillation using automatic external devices (AEDs). The electrocardiogram (ECG) signal's minute fluctuations in amplitude, duration, and shape make it difficult to categorize VAs from other types of arrhythmias. In this work, we implemented a nonlinear feature extraction-based approach to detect VAs. After the filtering and segmentation (8 seconds and 5 seconds) process, the ECG signals are decomposed using empirical mode decomposition (EMD). A new feature set is extracted from the decomposed intermediate mode functions (IMFs). The obtained features from the IMFs are fed to the support vector machine (SVM) classifier with a sequential minimal optimizer (SMO). We deployed the PhysioNet ECG databases to assess the model. We compare the proposed method to current practices using the performance metric provided by the average accuracy, sensitivity, and specificity values. Average accuracy levels increased on 5-second duration using the ML model instead of the more traditional approaches.