Atrial Fibrillation Detection Using Convolutional Neural Networks

Abstract

As part of the PhysioNet/Computing in Cardiology Challenge 2017, this work focuses on the classification of a single channel short electrocardiogram (ECG) signal into normal, atrial fibrillation (AF), others and noise classes. To this end, we propose a shallow convolutional neural network architecture which learns suitable features pertaining to each class while eliminating the need to extract the traditionally used ad hoc features. In particular, we first developed a robust R-peak detector and stacked sequence of fixed number of detected beats with R-peaks aligned. These stack of beats corresponding to a segment of ECG record are classified into one of the four aforementioned classes. To improve the robustness, multiple classifiers were trained to classify these segments. Overall record classification was then generated using an voting scheme from the classification results of individual segments. Our best submission result during the official phase has a score of 71% with F1 scores of 86%, 73% and 56% respectively for normal, AF and other classes respectively.

Classification of Atrial Fibrillation in Ecg Signal Using Deep Learning

“…In previous study Chandra and others [3] proposed beat stack classifier on 1-dimensional CNN model to classify stacks on R peak wave in MIT-BIH Arhythmia dataset.…”

“…However, diagnosing atrial fibrillation is not easy and can take a long time. This obstacle occurs because the rhythm of the ECG signal can change from an Atrial Fibrillation rhythm to a normal rhythm, and many rhythms that are not Atrial Fibrillation but have irregular RR intervals the same as Atrial Fibrillation [3].…”

Detection of Premature Ventricular Contractions Using Densely Connected Deep Convolutional Neural Network with Spatial Pyramid Pooling Layer

“…The network contains approximately 360000 trainable parameters. Our codes and checkpoint will be available upon acceptance 2…”

“…Some works also studied methods for the selection of handcraft features to improve detection performance [11,14,21].Various classifiers are used in the studies of PVC detection including artificial neural network [15,20,34],support vector machine [4] as well as clustering [3].Second, deep convolutional neural network (CNN)is showing advantages over traditional methods by providing a way of learning highly discriminative features automatically. Many works have studied its application in detecting abnormal heart beats including atrial fibrillation [2,27,30,7,25] and PVC [33,31,8] or other arrhythmia [22]. Currently, [33] achieves state-of-the-art results for PVC detection using a architecture combining multiple one-dimensional CNN and LSTM.However,all of the methods mentioned above are developed and tested using the same database and their crossdatabase generalization capability has not been fully validated.In clinic,however,the sampling rate of ECG data can be different from different devices.In such cases,its natural to think of training several networks for each specific ECG data which is time-consuming and sometimes impossible when the ECG data are limited.Therefore,its necessary to develop a generalized method that can maintain good performance across ECG data of varied sampling rates.In our study ,we propose a method based on densely connected convolutional neural network [10]and spatial pyramid pooling [13] for automatic PVC detection.The proposed network can take as input QRS complexes of arbitrary length and can be trained using multiple ECG databases with different sampling rates.Its cross-database generalization capability is verified on five open databases namely the MIT-BIH arrhythmia database,St-Petersburg Institute of Cardiological Technics 12-lead Arrhythmia Database,The MIT-BIH Normal Sinus Rhythm Database,The MIT-BIH Long Term Database and European ST-T Database.The performance on the MIT-BIH arrhythmia database which is a commonly used benchmark for arrhythmia detection is comparable to current state-of-the-art deep learning based method and our proposed network is much less complicated and easier to implement.…”

A Prototype Framework Design for Assisting the Detection of Atrial Fibrillation Using a Generic Low-Cost Biomedical Sensor

“…Some relevant algorithms for the detection of AF are the ones based on the analysis of the P-wave [10,11], or those that utilize a large set of features obtained from ECGs in an artificial neural network [12][13][14] or in a deep learning approach [15,16]. Furthermore, it is worth mentioning that several recent clinical trials on large populations focusing on AF detection have been carried out [17][18][19].…”