Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by diverse symptoms affecting social interaction, communication, and behavior. Diagnosing ASD is challenging due to variability among individuals and the lack of clear biomarkers. Although artificial intelligence (AI) has enhanced diagnostic accuracy, the "black-box" nature of many models limits their clinical use. Addressing current limitations, this paper introduces a multimodal ASD detection framework using deep neural networks (DNN) with explainable AI (xAI) to enhance model transparency. Our proposed approach achieves a mean 5-fold accuracy of 99.66% for the fMRI-based model and 99.83% for the multimodal-based model, which leverages both fMRI and phenotypic data, surpassing existing methods while offering better interpretability.