Enhancing Early Diagnosis of Autism Spectrum Disorder Using Multimodal Data and Explainable AI Models

Yeganeh Abdollahinejad, Md Faisal Kabir

Abstract: Autism Spectrum Disorder (ASD) presents profound challenges in early diagnosis due to its inherent complexity and variability. This research leverages a multimodal framework that integrates phenotypic data and neuroimaging quality metrics to establish a comprehensive machine learning pipeline for ASD prediction. Three machine learning models namely Gradient Boosting Machine (GBM), XGBoost, and Support Vector Machine (SVM) were trained and evaluated. Experimental results showed that GBM is best suited compared to other techniques for this case. To ensure clinical applicability, Shapley Additive Explanations (SHAP) were employed to elucidate feature contributions, fostering transparency and trust in the predictive process. This study highlights the potential of integrating machine learning models with interpretable frameworks to improve ASD diagnostics and support evidence-based clinical decision-making.