Brain-region specific autism prediction from electroencephalogram signals using graph convolution neural network.

Area of Science:

• Neuroscience
• Computational Neuroscience
• Biomedical Engineering

Background:

• Brain variations contribute to developmental disorders like autism spectrum disorder (ASD).
• Electroencephalography (EEG) signals offer valuable insights into brain function abnormalities for detecting neurological conditions.

Purpose of the Study:

• To investigate the efficacy of a Graph Convolutional Neural Network (GCNN) for predicting ASD.
• To analyze EEG data from autistic and typically developing children to identify neurological markers of ASD.

Main Methods:

• EEG data were collected from 8 autistic and 8 typically developing children.
• A GCNN model was employed for ASD prediction following autoregressive and spectral feature extraction.
• EEG data utilized 257 channels, with 71 channels (10-10 international equivalents) analyzed across 12 brain regions.

Main Results:

• The anterior-frontal brain region demonstrated the highest predictive capability for ASD.
• The GCNN model achieved an accuracy of 87.07% in ASD prediction.
• This highlights the GCNN method's suitability for EEG-based ASD detection.

Conclusions:

• The study provides a detailed dataset that deepens the understanding of the neurological underpinnings of ASD.
• Findings can assist healthcare practitioners in the diagnosis of ASD.
• The GCNN approach shows potential for improving ASD detection through EEG analysis.