SleepFCN: A Fully Convolutional Deep Learning Framework for Sleep Stage Classification Using Single-Channel...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Artificial Intelligence

Background:

• Sleep stage classification is crucial for assessing sleep quality and diagnosing sleep disorders.
• Accurate classification relies on analyzing electroencephalogram (EEG) signals.
• Existing methods may face challenges with data imbalance and computational efficiency.

Purpose of the Study:

• To introduce a novel fully convolutional neural network (FCNN) architecture named SleepFCN.
• To classify sleep stages into five classes using single-channel EEGs.
• To improve the accuracy and learning speed of sleep stage classification.

Main Methods:

• Developed SleepFCN, a fully convolutional neural network architecture.
• Incorporated multi-scale feature extraction (MSFE) and residual dilated causal convolutions (ResDC) for feature extraction and temporal encoding.
• Utilized convolutional layers with 1-sized kernels and addressed class imbalance in the loss function.

Main Results:

• SleepFCN demonstrated superior performance compared to state-of-the-art methods.
• The model achieved high classification correctness for sleep stages.
• SleepFCN exhibited a faster learning speed in experimental evaluations.

Conclusions:

• SleepFCN offers a powerful and efficient approach for automated sleep stage classification using single-channel EEGs.
• The proposed architecture effectively handles the complexities of sleep EEG data.
• This method has the potential to enhance the diagnosis and management of sleep disorders.