Classificação de ECG utilizando um sistema de inferência neuro-fuzzy adaptativo

Abstract

This monograph presents a study and implementation of an intelligent system capable of classifying electrocardiogram (ECG) signals using an Adaptive Neuro-Fuzzy Inference System (ANFIS). Additionally, cardiac signals are a rich source of information about heart health, allowing for the analysis of its clinical conditions. To make this possible, electrocardiography emerges as a crucial technique, enabling the non-invasive recording of the heart's electrical activity through electrodes. From the extraction of these signals, there is room for the application of advanced signal processing and data analysis techniques, enhancing the diagnosis of heart diseases. Due to the availability of ECG signal databases, the intelligent system can learn and perform functions similar to those of cardiologists. The detection of cardiac abnormalities is carried out through an ANFIS, pre-processed by subtractive clustering, enabling precise classification of ECG signals. The central objective is to identify crucial characteristics in ECG signals, such as heart rate variation, to determine whether the patient's heartbeat is within normal ranges or if there are irregularities. Furthermore, five types of heartbeats were selected for classification: normal sinus rhythm, atrial premature contraction (APC), left bundle branch block (LBBB), right bundle branch block (RBBB), and paced beat (PB). The results obtained in this monograph demonstrate an average accuracy of 98.27%, an average sensitivity of 95.68%, and an average specificity of 98.92%, results comparable to Artificial Neural Network (ANN) algorithms, reinforcing the effectiveness of the proposed system.