Detection of Cardiac Arrhythmia using Autonomic Nervous System, Gaussian Mixture Model and Artificial Neural Network

In this study, a new technique which detects anomalies in skin sympathetic nerve activity (SKNA) by using state-of-the-art signal processing and machine learning methods is developed to perform the robust detection of cardiac arrhythmia (CA). For this purpose, a signal processing technique that simultaneously obtains SKNA and ECG from wideband recordings on MIT-BIH database is developed. By using preprocessed data, a novel feature extraction technique which obtains SKNA features that are critical for the reliable detection of CA is developed. By using extracted features, a supervised learning technique based on artificial neural network (ANN) and an unsupervised learning technique based on Gaussian mixture model (GMM) are developed to perform the robust detection of SKNA anomalies. A Neyman-Pearson type of approach is developed to perform the robust detection of outliers that correspond to CA. The performance results of the proposed technique over MIT-BIH database showed that the technique provides highly reliable detection of CA by performing the robust detection of SKNA anomalies. Therefore, in cases where the diagnostic information of ECG is not sufficient for the reliable diagnosis of CA, the proposed technique can provide early diagnosis of the disease, which can lead to a significant reduction in the mortality rates of cardiovascular diseases.