Objective:

To investigate the ability of machine learning-based classification models using multimodal neuroimaging data to identify brain characteristics associated with autism across different developmental stages.

Approach:

• Participants: 144 participants aged 5 to 18 years with structural MRI (sMRI), diffusion MRI (dMRI), and resting-state functional MRI (rs-fMRI) data.
• Data Source: Data obtained from the Autism Brain Imaging Data Exchange (ABIDE) database.
• Methodology: Radiomic features were extracted and used to train support vector machine (SVM) classifiers for neuroimaging pattern identification.
• Age Groups: Participants were divided into three age sub-groups: younger children (5–11 years), adolescents (12–18 years), and the entire cohort (5–18 years).
• Model Evaluation: Model performance was evaluated using leave-one-out cross-validation across 30 diagnosis-balanced data splits.
• Feature Importance: Feature-importance analyses were conducted to identify significant neuroimaging features for classification.

Key Findings:

• Classification accuracies for unimodal models ranged from 68.3% to 75.3% for sMRI, 69.3% to 77.6% for dMRI, and 66.3% to 69.9% for rs-fMRI.
• dMRI showed the highest performance with a 77.6% accuracy in younger children (5–11 years).
• The multimodal approach improved classification performance, achieving accuracies of 78.9%, 76.7%, and 70.5% in younger, adolescent, and entire cohorts, respectively.
• The most informative brain regions for classification differed between children and adolescents.
• Several diffusion-derived features correlated with social responsiveness scores.

Interpretation:

The study demonstrates the potential of multimodal neuroimaging-based machine learning models to identify development-specific biomarkers associated with autism.

Conclusion:

The findings emphasize the importance of integrating age-stratified analyses of multimodal neuroimaging to capture autism-associated developmental brain characteristics.