ULW-SleepNet: An Ultra-Lightweight Network for Multimodal Sleep Stage Scoring
arXiv:2602.23852v1 Announce Type: new Abstract: Automatic sleep stage scoring is crucial for the diagnosis and treatment of sleep disorders. Although deep learning models have advanced the field, many existing models are computationally demanding and designed for single-channel electroencephalography (EEG), limiting their practicality for multimodal polysomnography (PSG) data. To overcome this, we propose ULW-SleepNet, an ultra-lightweight multimodal sleep stage scoring framework that efficiently integrates information from multiple physiological signals. ULW-SleepNet incorporates a novel Dual-Stream Separable Convolution (DSSC) Block, depthwise separable convolutions, channel-wise parameter sharing, and global average pooling to reduce computational overhead while maintaining competitive accuracy. Evaluated on the Sleep-EDF-20 and Sleep-EDF-78 datasets, ULW-SleepNet achieves accuracies of 86.9% and 81.4%, respectively, with only 13.3K parameters and 7.89M FLOPs. Compared to state-of-
arXiv:2602.23852v1 Announce Type: new Abstract: Automatic sleep stage scoring is crucial for the diagnosis and treatment of sleep disorders. Although deep learning models have advanced the field, many existing models are computationally demanding and designed for single-channel electroencephalography (EEG), limiting their practicality for multimodal polysomnography (PSG) data. To overcome this, we propose ULW-SleepNet, an ultra-lightweight multimodal sleep stage scoring framework that efficiently integrates information from multiple physiological signals. ULW-SleepNet incorporates a novel Dual-Stream Separable Convolution (DSSC) Block, depthwise separable convolutions, channel-wise parameter sharing, and global average pooling to reduce computational overhead while maintaining competitive accuracy. Evaluated on the Sleep-EDF-20 and Sleep-EDF-78 datasets, ULW-SleepNet achieves accuracies of 86.9% and 81.4%, respectively, with only 13.3K parameters and 7.89M FLOPs. Compared to state-of-the-art methods, our model reduces parameters by up to 98.6% with only marginal performance loss, demonstrating its strong potential for real-time sleep monitoring on wearable and IoT devices. The source code for this study is publicly available at https://github.com/wzw999/ULW-SLEEPNET.
Executive Summary
The proposed ULW-SleepNet framework achieves high accuracy in multimodal sleep stage scoring with significantly reduced computational overhead. It incorporates novel techniques such as Dual-Stream Separable Convolution (DSSC) Block and channel-wise parameter sharing, resulting in a model with only 13.3K parameters and 7.89M FLOPs. Evaluated on Sleep-EDF-20 and Sleep-EDF-78 datasets, ULW-SleepNet achieves accuracies of 86.9% and 81.4%, respectively, demonstrating its potential for real-time sleep monitoring on wearable and IoT devices.
Key Points
- ▸ Ultra-lightweight multimodal sleep stage scoring framework
- ▸ Novel Dual-Stream Separable Convolution (DSSC) Block and channel-wise parameter sharing
- ▸ Significant reduction in computational overhead with minimal performance loss
Merits
Efficient Architecture
The proposed ULW-SleepNet framework efficiently integrates information from multiple physiological signals, reducing computational overhead while maintaining competitive accuracy.
Demerits
Limited Generalizability
The model's performance may not generalize well to other datasets or populations, and further evaluation is needed to assess its robustness.
Expert Commentary
The ULW-SleepNet framework represents a significant advancement in multimodal sleep stage scoring, demonstrating the potential for ultra-lightweight models to achieve high accuracy with minimal computational overhead. The use of novel techniques such as DSSC Block and channel-wise parameter sharing enables efficient integration of information from multiple physiological signals. However, further evaluation is needed to assess the model's robustness and generalizability to other datasets and populations.
Recommendations
- ✓ Further evaluation of ULW-SleepNet on diverse datasets and populations to assess its robustness and generalizability
- ✓ Exploration of applications in real-time sleep monitoring on wearable and IoT devices
