Abstract:Sleep staging serves as a fundamental assessment for sleep quality measurement and sleep disorder diagnosis. Although current deep learning approaches have successfully integrated multimodal sleep signals, enhancing the accuracy of automatic sleep staging, certain challenges remain, as follows: 1) optimizing the utilization of multi-modal information complementarity, 2) effectively extracting both long- and short-range temporal features of sleep information, and 3) addressing the class imbalance problem in sleep data. To address these challenges, this paper proposes a two-stream encode-decoder network, named TSEDSleepNet, which is inspired by the depth sensitive attention and automatic multi-modal fusion (DSA2F) framework. In TSEDSleepNet, a two-stream encoder is used to extract the multiscale features of electrooculogram (EOG) and electroencephalogram (EEG) signals. And a self-attention mechanism is utilized to fuse the multiscale features, generating multi-modal saliency features. Subsequently, the coarser-scale construction module (CSCM) is adopted to extract and construct multi-resolution features from the multiscale features and the salient features. Thereafter, a Transformer module is applied to capture both long- and short-range temporal features from the multi-resolution features. Finally, the long- and short-range temporal features are restored with low-layer details and mapped to the predicted classification results. Additionally, the Lovász loss function is applied to alleviate the class imbalance problem in sleep datasets. Our proposed method was tested on the Sleep-EDF-39 and Sleep-EDF-153 datasets, and it achieved classification accuracies of 88.9% and 85.2% and Macro-F1 scores of 84.8% and 79.7%, respectively, thus outperforming conventional traditional baseline models. These results highlight the efficacy of the proposed method in fusing multi-modal information. This method has potential for application as an adjunct tool for diagnosing sleep disorders.

Dreaming is All You Need

CareSleepNet: A Hybrid Deep Learning Network for Automatic Sleep Staging

Domain Invariant Representation Learning and Sleep Dynamics Modeling for Automatic Sleep Staging

MSF-SleepNet: Multi-Stream Fusion Network with Contrastive Learning for Sleep Stage Classification.

NeuroNet: A Novel Hybrid Self-Supervised Learning Framework for Sleep Stage Classification Using Single-Channel EEG

Explore intrinsic geometry of sleep dynamics and predict sleep stage by unsupervised learning techniques

Interpreting Sleep Activity Through Neural Contrastive Learning

WaveSleepNet: An Interpretable Network for Expert-like Sleep Staging

A Deep Learning Approach with an Attention Mechanism for Automatic Sleep Stage Classification

S4Sleep: Elucidating the design space of deep-learning-based sleep stage classification models

Intelligent automatic sleep staging model based on CNN and LSTM

SLEEPNET: Automated Sleep Staging System via Deep Learning

Deep Learning in Automatic Sleep Staging With a Single Channel Electroencephalography

Multi-Modal Sleep Stage Classification With Two-Stream Encoder-Decoder

XSleepNet: Multi-View Sequential Model for Automatic Sleep Staging

LGSleepNet: an Automatic Sleep Staging Model Based on Local and Global Representation Learning.

A Dual-Stream Deep Neural Network Integrated with Adaptive Boosting for Sleep Staging

RecSleepNet: an Automatic Sleep Staging Model Based on Feature Reconstruction

Single Channel Sleep Staging Based on Unsupervised Feature Learning

A Two-Stage Neural Network for Sleep Stage Classification Based on Feature Learning, Sequence Learning, and Data Augmentation

The overfitted brain: Dreams evolved to assist generalization