Deep Learning Detection of Corrupted Segments in Recordings from Wearable Devices to Improve Atrial Fibrillation Screening
Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia in clinical practice. It is associated with an increased risk of cardiovascular events, but its early detection is an unresolved challenge. For that purpose, long-term wearable electrocardiogram (ECG) recording systems are being widely used in the last years, because the arrhythmia often starts with asymptomatic and very short episodes. However, these equipments work in highly dynamics and ever-changing environments, thus providing ECG signals strongly corrupted with different kinds of noises. In this context, ECG quality assessment results essential for a precise and robust AF detection. Hence, this work introduces a deep learning-based algorithm to discern between high- and low-quality segments in single-lead ECG recordings obtained from patients with intermittent AF. The method is based on the high learning capability of a convolutional neural network, which has been trained with 2-D images obtained when turning ECG signals into wavelet scalograms. The obtained results have reported a great ability to discern between high- and low-quality ECG excerpts about 95%, only misclassifying around 6% of clean AF intervals as noisy segments. These outcomes have improved by more than 20% performances of most previous ECG quality assessment algorithms also dealing with AF signals.
Paper
Full text
Deep Learning Detection of Corrupted Segments in Recordings from Wearable Devices to Improve Atrial Fibrillation Screening
Semantic Scholar · Medicine · 2020
Abstract
Atrial fibrillation (AF) is the most common sustained cardiac arrhythmia in clinical practice. It is associated with an increased risk of cardiovascular events, but its early detection is an unresolved challenge. For that purpose, long-term wearable electrocardiogram (ECG) recording systems are being widely used in the last years, because the arrhythmia often starts with asymptomatic and very short episodes. However, these equipments work in highly dynamics and ever-changing environments, thus providing ECG signals strongly corrupted with different kinds of noises. In this context, ECG quality assessment results essential for a precise and robust AF detection. Hence, this work introduces a deep learning-based algorithm to discern between high- and low-quality segments in single-lead ECG recordings obtained from patients with intermittent AF. The method is based on the high learning capability of a convolutional neural network, which has been trained with 2-D images obtained when turning ECG signals into wavelet scalograms. The obtained results have reported a great ability to discern between high- and low-quality ECG excerpts about 95%, only misclassifying around 6% of clean AF intervals as noisy segments. These outcomes have improved by more than 20% performances of most previous ECG quality assessment algorithms also dealing with AF signals.