G2-ResNeXt: a novel model for ECG signal classification

Electrocardiograms (ECG) are the primary basis for diagnosis of cardiovascular diseases. However, due to large volume patients’ ECG data, manual is time-consuming and laborious. Therefore, intelligent automatic signal classification an important technique overcoming shortage medical resources. This paper proposes a novel model inter-patient heartbeat classification, named G2-ResNeXt, which adds two-fold grouping convolution (G2) original ResNeXt structure, as achieve better feature extraction signals. Experiments, conducted on MIT-BIH arrhythmia database, confirm that proposed outperforms all state-of-the-art models considered (except GRNN one classes), by achieving overall accuracy 96.16%, xmlns:xlink="http://www.w3.org/1999/xlink">sensitivity xmlns:xlink="http://www.w3.org/1999/xlink">precision 97.09% 95.90%, respectively, ventricular ectopic heartbeats (VEB), 80.59% 82.26%, supraventricular (SVEB).