Multimodal MRI machine learning identifies age-specific brain features in autistic youth

Multimodal MRI (sMRI, dMRI, rs-fMRI) machine learning classifier achieved 78.9% accuracy for autism in children 5-11 years, outperforming single-modality models. Diffusion features correlated with social responsiveness scores.

• n=144 participants aged 5-18 from ABIDE; leave-one-out cross-validation with 30 diagnosis-balanced splits.
• Multimodal SVM classifier accuracy: 78.9% in children 5-11y, 76.7% in adolescents 12-18y, 70.5% in full cohort; dMRI alone best unimodal at 77.6% in younger children.
• Diffusion-derived features significantly correlated with social responsiveness scores, highlighting the relevance of microstructural white/gray matter.

Automated summary

RadPigeon summaries are original and for information only. They are not clinical advice.