Detection of Ventricular Fibrillation by combining Signal Processing and Machine Learning approach

Ventricular Fibrillation is a potentially fatal cardiac disorder that occurs when electrical impulses in the ventricles are disrupted, causing the heart to quiver instead of pump. In order to preserve lives during this form of arrhythmia, a strong current impulse is passed. Electrocardiograms (ECGs) record the electrical activity of the human heart, and specialists with years of experience may interpret the ECG signal to determine the heart's condition. Since it is a life-threatening disease, its earlier detection and prevention can help survive a patient's life. The fundamental idea behind tackling this challenge was to create an algorithm that could identify trends from continuous ECG readings from various individuals and identify arrhythmias early on. An efficient data was built for classification utilizing a Random Forest classifier algorithm employing signal processing tools such as Empirical Mode Decomposition (EMD) and Discrete Fourier Transform (DFT) for feature extraction. The pre-processed data when fed into the proposed machine learning method results in an accuracy of 96.58% and two classes were classified correctly with equal confidence (Specificity = 94.26% and Sensitivity = 98.97%). Furthermore, the results are compared with various other machine learning classification algorithms like Logistic Regression, Decision Tree classifier, Extra tree classifier where the accuracy was 86.49%, 91.77%, 95.84% respectively. The results obtained after experimental validation of proposed Random Forest classifier algorithm against the other machine learning achieves highest accuracy with optimal specificity and sensitivity.