ECG Pattern Recognition Technique for Atrial Fibrillation Detection

Atrial Fibrillation (AF) is the most common pathologic of sinus tachycardia, which is the result of an increased rate of depolarization in the sinoatrial node (the sinoatrial node discharges electrical impulses at a higher frequency than normal). In this light, its detection at an early stage is essential for treatments prescription. In this context, we propose an artificial neural network (ANN) architecture using ECG patterns to perform the AF detection. ECG signals are classified into three classes (Normal Sinus Rhythm, abnormal signal with AF, and noisy ECG signals). The proposed technique has been implemented on two types of databases (MIT-BIH database and processed MIT-BIH) using two different experiments. Data segments of 10 seconds length have been used. The achieved experimental results proved that the proposed ANN technique has excellent accuracy results without the need for feature extraction to reduce information parameters. Our work has surpassed the state of the art in terms of specificity, precision, and accuracy. Therefore, we enable clinicians to detect automatically the patients with AF disease.