An end-to-end 3D convolutional neural network for decoding attentive mental state.

Area of Science:

• Neuroscience
• Machine Learning
• Biomedical Engineering

Background:

• Accurate detection of attentive mental states is crucial for neurofeedback and treating Attention Deficit and Hyperactivity Disorder (ADHD).
• Current electroencephalogram (EEG) detection methods face challenges in effectively representing temporal and spatial data.
• Deep learning (DL) approaches show promise in brain-computer interface (BCI) research.

Purpose of the Study:

• To propose a novel 3D representation for EEG signals to improve attention detection.
• To develop and evaluate a 3D convolutional neural network (CNN) model for enhanced attention detection.

Main Methods:

• Introduced a 3D representation of EEG signals to capture temporal and spatial topological characteristics.
• Developed a 3D CNN model with cascade and parallel convolution operations for multi-scale feature extraction.
• Evaluated the model on a public dataset involving twenty-six subjects across different classification scenarios.

Main Results:

• The proposed 3D CNN model demonstrated superior performance compared to baseline methods.
• The model achieved better results in intra-subject, inter-subject, and subject-adaptive classification.
• The 3D representation effectively preserved crucial temporal and spatial information from EEG data.

Conclusions:

• The 3D CNN model shows significant potential for accurate attentive mental state detection.
• This approach offers a promising advancement for neurofeedback and ADHD treatment strategies.
• The study highlights the efficacy of 3D CNNs in analyzing complex EEG data for cognitive state monitoring.