Simultaneous Sleep Stage and Sleep Disorder Detection from Multimodal Sensors Using Deep Learning.

Area of Science:

• Biomedical Engineering
• Computational Neuroscience
• Sleep Medicine

Background:

• Sleep scoring is crucial for diagnosing sleep disorders.
• Simultaneous identification of sleep stages and disorders improves diagnostic accuracy.
• Current methods often rely on single modalities or labels, limiting performance.

Purpose of the Study:

• To investigate the automatic recognition of sleep stages and disorders from multimodal sensory data.
• To propose a novel distributed multimodal and multilabel decision-making system (MML-DMS).
• To enhance diagnostic performance by fusing multimodal and multilabel information.

Main Methods:

• Utilized multimodal sensory data including electroencephalogram (EEG), electrocardiogram (ECG), and electromyogram (EMG).
• Developed a distributed multimodal and multilabel decision-making system (MML-DMS) integrating deep convolutional neural networks (CNNs) and shallow neural networks (NNs).
• Employed VGG16 CNN structures and tested on the PhysioNet CAP Sleep Database.

Main Results:

• Achieved an average classification accuracy of 94.34% and F1 score of 0.92 for six sleep stages.
• Attained an average classification accuracy of 99.09% and F1 score of 0.99 for eight sleep disorders.
• Demonstrated superior diagnostic performance compared to single-label and single-modality approaches.

Conclusions:

• The proposed MML-DMS significantly improves the accuracy of sleep stage and disorder detection.
• Fused multimodal and multilabel approaches offer a more robust diagnostic tool for sleep disorders.
• The findings represent a significant advancement over existing state-of-the-art methods in automatic sleep analysis.