DSD-R: Deep Learning Based Segmentation for the Detection of R peaks in ECG Signals
Küçük Resim Yok
Tarih
2022
Dergi Başlığı
Dergi ISSN
Cilt Başlığı
Yayıncı
Ieee
Erişim Hakkı
info:eu-repo/semantics/closedAccess
Özet
Electrocardiography (ECG) is the recording of the electrical activity of the heart through electrodes. ECG signals are crucial in the early diagnosis of numerous cardiac diseases. Therefore, it is very important to read and analyze these signals using state-of-the-art technologies. The regular wave shapes in ECG data are frequently disturbed when certain heart diseases occur and these changes in signals help for detecting the disease. Signal processing and machine learning-based methods are widely used for this purpose. In recent years, deep learning-based methods have become widespread, and they offer promising results. This study aims to segmentation-based detection of R-peak locations in ECG signals. First, the ECG signal is transformed into a Continuous Wavelet Transform (CWT) based scalogram image, and then U-Net-based deep learning architectures are utilized for the segmentation. The comparisons are carried out on MIT-BIH Arrhythmia Database (MIT-DB). Whereas all methods provide promising results, U-Net 3+ model achieves 0.99 in Precision, 0.98 in Recall, 0.99 in F1 score, and 0.98 in Accuracy with the lowest parameter size.
Açıklama
Medical Technologies Congress (TIPTEKNO) -- OCT 31-NOV 02, 2022 -- Antalya, TURKEY
Anahtar Kelimeler
ECG, Scalogram, Segmentation
