BGcsSENet: Bidirectional GRU With Spatial and Channel Squeeze-Excitation Network for Bundle Branch Block Detection

Left bundle branch block (LBBB) and right bundle branch block (RBBB) are the common but neglected arrhythmias clinically. However, the conventional LBBB and RBBB diagnosis methods are strenuous and time-consuming because this process relies heavily on the physician's visual observation of elect...

Full description

Saved in:
Bibliographic Details
Published in:IEEE transactions on human-machine systems 2023-04, Vol.53 (2), p.1-9
Main Author: Wang, Jibin
Format: Article
Language:English
Subjects:
Bidirectional gated recurrent unit (B-GRU)
Computer architecture
Effectiveness
electrocardiogram
Electrocardiography
Excitation
Feature extraction
Modules
spatial and channel squeeze-excitation network (scSENet)
Testing
Training
Visual observation
Wavelet transforms
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Left bundle branch block (LBBB) and right bundle branch block (RBBB) are the common but neglected arrhythmias clinically. However, the conventional LBBB and RBBB diagnosis methods are strenuous and time-consuming because this process relies heavily on the physician's visual observation of electrocardiogram. In this article, we developed a novel network module based on the existing bidirectional gated recurrent unit (B-GRU) network for automated LBBB, RBBB, and normal sinus rhythm detection. Inspired from spatial and channel squeeze-excitation network (scSENet), the proposed module is capable of effectively recalibrating feature representation in B-GRU using both channel squeeze-excitation and spatial squeeze-excitation modules along channel and space to emphasize useful features while suppressing weak ones, namely, BGcsSENet. To further verify its effectiveness, we leverage the standard GRU and B-GRU modules as control models across the two public databases. Extensive experiments confirm the effectiveness of BGcsSENet, while yielding better performance than existing GRU, B-GRU modules, and many published approaches. Particularly, our BGcsSENet model is the first research to refine the network architecture of B-GRU, thus demonstrating considerable application potential on lightweight devices.
ISSN:2168-2291
2168-2305
DOI:10.1109/THMS.2023.3244938