GM-VGG-Net: A Gray Matter-Based Deep Learning Network for Autism Classification.

Area of Science:

• Neuroimaging
• Artificial Intelligence
• Medical Diagnostics

Background:

• Autism Spectrum Disorder (ASD) affects approximately 1 in 59 individuals.
• Current ASD diagnostic methods often rely on functional brain regions or multi-tissue AI models.
• Structural magnetic resonance imaging (MRI) offers a potential avenue for improved diagnostic tools.

Purpose of the Study:

• To develop an efficient deep learning network for identifying ASD.
• To utilize whole brain gray matter (GM) tissues from structural MRI for ASD diagnosis.
• To establish a novel diagnostic approach using VGG-Net architecture.

Main Methods:

• A VGG-based deep learning network was developed.
• The network was trained using 132 MRI T1 images from normal controls and 140 from ASD patients.
• Data was sourced from the Autism Brain Imaging Data Exchange (ABIDE) dataset.

Main Results:

• The deep learning model achieved 97% training accuracy and 96% validation accuracy.
• The model demonstrated effective performance over 50 epochs without overfitting.
• No statistically significant difference in participant numbers or age was observed between ASD and control groups.

Conclusions:

• This study presents the first application of VGG-Net using solely gray matter tissue for ASD diagnosis.
• The findings suggest the potential of deep learning models with structural MRI for accurate ASD identification.
• This approach offers a promising, efficient, and potentially more accessible method for ASD diagnosis.