Multiple Classification of Brain MRI Autism Spectrum Disorder by Age and Gender Using Deep Learning.

Area of Science:

• Neuroscience
• Artificial Intelligence
• Medical Imaging

Background:

• Current limitations exist in the rapid and definitive diagnosis and treatment of Autism Spectrum Disorders (ASD).
• Novel technological solutions are needed to improve ASD diagnosis.
• Investigating the influence of demographic factors like age and gender on ASD diagnosis is crucial.

Purpose of the Study:

• To develop and evaluate a deep learning (DL) system for multi-classification of ASD based on age and gender.
• To determine the contribution of age and gender factors to ASD diagnosis.
• To compare the performance of custom DL models against pre-trained models using transfer learning.

Main Methods:

• Structural MRI (sMRI) scans from individuals with ASD and Typical Development (TD) were pre-processed.
• Canny Edge Detection (CED) and data augmentation (DA) techniques were employed for data preparation.
• Three Convolutional Neural Network (CNN) models were developed using Grid Search Optimization (GSO): a gender-based quadruple classification, an age-based quadruple classification, and a combined age and gender octal classification.
• The models were validated using five-fold cross-validation and compared with pre-trained models via transfer learning.

Main Results:

• The gender-based model achieved 80.94% accuracy.
• The age-based model achieved 85.42% accuracy.
• The combined age and gender model achieved 67.94% accuracy, outperforming pre-trained models in diagnostic accuracy for ASD when considering these factors.

Conclusions:

• Age and gender are effective factors influencing the diagnosis of ASD.
• The developed DL system demonstrates the potential for improved ASD diagnosis through multi-classification strategies.
• Custom-designed DL models incorporating demographic factors show promise compared to standard pre-trained models.