غير محددة في المصدر 2020 بحاجة إلى مراجعة

A Deep Learning (Neural Network) Approach for ECG Heartbeat Classification: Arrhythmia Detection

الملخص

Nearly 17 million people die every year due to cardiovascular disease according to World Health Organization and one of the diseases associated with the human heart is arrhythmia.Arrhythmia is the irregularity and abnormality of heartbeat rhythm.To detect arrhythmia, ECG (Electrocardiogram) is used for diagnosing it, but in order to interpret well ECG signals, many automated analyses method have been proposed over the years but not all gave a satisfactory result.However, Deep learning especially Convolutional Neural Network has been successfully used for the detection and classification of ECG signal because it is an end to end model, which means it does not require pre-processing before network training except normalizing the dataset.In this thesis, a 1D-Convolutional Neural Network model (1D-CNN) was proposed, which is mainly used for signal and time-series data.A total of 7636 consisting of 5 different classes of arrhythmias were extracted which includes Normal sinus rhythm n, Atrial Premature beat, Premature ventricular contraction, Atrial fibrillation, and Atrial flutter from the MIT -BIH Arrhythmia database.By experimenting with 5 different 1D-CNN layered model, different results were achieved.The results obtained all showed that increasing or decreasing the number of layers with respect to changing the values of hyperparameters can affect performance results of a network. .

الكلمات المفتاحية

غير متوفر

التخصصات

رسائل وأطاريح مشابهة