A Sequential End-to-End Neonatal Sleep Staging Model with Squeeze and Excitation Blocks and Sequential Multi-Scale...

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Pediatric Sleep Medicine

Background:

• Automatic sleep staging in neonates is crucial for objective assessment.
• Existing methods often lack temporal information or suffer from high computational costs and ambiguity.
• This limits their applicability across diverse clinical scenarios.

Purpose of the Study:

• To propose a novel sequential end-to-end sleep staging model, SeqEESleepNet.
• To enable parallel processing of sequential epochs for efficient temporal information extraction.
• To develop a fast-training model adaptable to various neonatal sleep staging scenarios.

Main Methods:

• SeqEESleepNet integrates a sequence epoch generation (SEG) module, a sequential multi-scale convolution neural network (SMSCNN), and squeeze and excitation (SE) blocks.
• The SEG module converts independent epochs into sequential signals to capture inter-epoch temporal dynamics.
• SMSCNN extracts multi-scale and temporal features, while SE blocks optimize feature weighting.

Main Results:

• SeqEESleepNet achieved 71.8% overall accuracy, 71.8% F1-score, and a 0.684 Kappa coefficient on a three-class classification task using single-channel EEG.
• The proposed method outperformed existing state-of-the-art approaches in clinical dataset evaluation.
• The model demonstrated efficient training and parallel processing capabilities.

Conclusions:

• SeqEESleepNet provides an effective and computationally efficient solution for neonatal sleep staging.
• The model's ability to process sequential epochs enhances temporal information utilization.
• This approach holds promise for developing convenient, multi-scenario neonatal sleep staging tools.