A Decision Support System Based on multi-head convolutional and Recurrent Neural Networks for assisting physicians in...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Attention-Deficit Hyperactivity Disorder (ADHD) is a prevalent neurodevelopmental disorder in youth.
• Current ADHD diagnostic methods are subjective and time-consuming.
• Objective biomarkers are needed for accurate ADHD detection.

Purpose of the Study:

• To propose MCRNet, a novel Multi-head Convolutional and Recurrent Neural Network.
• To utilize electroencephalography (EEG) and deep learning (DL) for ADHD diagnosis.
• To develop an objective and reliable tool for ADHD detection.

Main Methods:

• MCRNet employs a parallel architecture with convolutional and recurrent neural network modules.
• A two-stage multi-head approach enhances feature extraction within the convolutional module.
• Cross-subject validation was used to ensure generalizability to new patients.

Main Results:

• MCRNet achieved 94.87% accuracy and 98.33% recall in identifying ADHD cases.
• The model demonstrated superior performance compared to existing methods using raw EEG signals.
• Results indicate MCRNet's high reliability for objective ADHD diagnosis.

Conclusions:

• MCRNet shows significant potential for aiding ADHD diagnosis.
• The model's architecture improves feature extraction and classification from EEG data.
• Future research should explore MCRNet's explainability and validate on diverse EEG datasets.