Detection of Atrial Fibrillation by Custom Designed Convolutional Neural Network: A Medical Viewpoint
Bildiriler > Tam Metin Bildiri
Yazarlar: Yasmin Cetintemel, Eda Nur Oner, Cagla Sarvan, Nalan Ozkurt, Evrim ŞİMŞEK
2022 Innovations in Intelligent Systems and Applications Conference (ASYU), , , Uluslararası, Tam Metin Bildiri, cilt 2, ss 1-6
DOI: 10.1109/asyu56188.2022.9925304
Özet
In this study, a specially designed convolutional neural network (CNN) model was used to detect atrial fibrillation (AF) by electrocardiographic (ECG) signals. For the classification process, ECG signals were obtained from the “The PhysioNet/Computing in Cardiology Challenge 2017” database. Prior information is always useful for the construction of an artificial intelligence model. In this problem, the experts denote that the absence of P wave and irregular R-R intervals indicate AF. Thus, in this study, regarding this valuable information, the average of the R-R intervals and the length of P waves were used as filter lengths of the first convolutional layers. It has been observed that selecting filter lengths with prior information contributes to creating a faster learning model with higher accuracy than the classification made with randomly selected filter sizes and numbers.