.png){kind=icon}
.png){kind=icon}
Return
Multi-channel EEG based layered sleep spindles detection algorithm using dual-branch BiLSTM
.png){kind=icon}
PRE
AI
D
Z
T
F
X
J
DOI：10.1016/j.bspc.2026.111225.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Sleep spindles, as key physiological signals during sleep, exhibit a spindle shape in electroencephalogram (EEG) and play a crucial role in cognition and memory consolidation. Existing research has shown that sleep spindles, as localized brain phenomena, exhibit synchronization across multiple EEG channels and concentrate in specific brain regions. This study aims to develop a layered sleep spindle detection algorithm based on multi-channel EEG with temporal dynamic characteristics. First, it performs feature-based threshold detection and result encoding on each channel. Then, it employs an Extreme Gradient Boosting (XGBoost) ensemble learning model that integrates multichannel spatio-temporal correlation features to generate preliminary detection results by constructing hierarchical feature interaction mechanisms. Finally, addressing the significant differences in sleep spindle characteristics between Benign Epilepsy with Centro Temporal Spikes (BECTS) patients and normal states, a dual-branch bidirectional long short-term memory (BiLSTM) network is utilized for in-depth detection. Experiments were conducted on 17 BECTS patients in the Children’s Hospital, Zhejiang University School of Medicine (CHZU). The proposed algorithm achieves the precise capture of localized features, thereby enhancing the accuracy of detection results. With an overall detection accuracy of 90.5%, the recall of 91.0%, the precision of 87.4%, and the F1 score of 89.2%, the results confirm the effectiveness and clinical applicability of this method.
Journal
IF:
4.9
Papers:
9.7K
Citations:
2.4W
Organization
Cited Papers