Multimodal classification for autism spectrum disorder using artificial intelligence

Abstract

Autism Spectrum Disorder (ASD) is a multifaceted phenomenon that emerges from early brain development. It is characterized by a constellation of symptoms that vary widely among individuals, forming a unique mosaic of challenges and strengths. Challenges in social interaction, communication, and repetitive behaviors newlinecharacterize ASD. Core concerns include a wide symptomatic variation, comorbid nature, self-injurious characteristics, and the newlinesubstantial need for care. According to the latest reports released by the World Health Organization and the Centers for Disease Control and Prevention (CDC), it is estimated to affect 1 child in 54 worldwide, which is approximately 1% of children of the global newlinepopulation. The prevalence rate for the 8 years and below age group reported in 2023 is 1: 36, which has increased by 178% since 2000 newlinein the United States and is increasing worldwide like a crisis. The ratio of male autistic cases is 4 times of female cases. Based on the severity of the disorder, autism is divided into three levels. The most severe condition reported in a census was self-injurious behavior with a low IQ level (IQlt50). The exponential increase in the growth rate of autism among cognitive disorders raises significant concerns and urges an early diagnosis. Recent developments in cognitive disorders include harnessing Artificial Intelligence (AI) and highresolution imaging modalities like structural-Magnetic Resonance Imaging (sMRI), resting-state functional MRI (rs-fMRI), and newlineDiffusion Tensor Imaging (DTI). While clinical findings from images provide crucial insights but are inaccessible to traditional newlinemethods, advanced AI algorithms hold the potential to revolutionize healthcare by enabling earlier diagnosis. The proposed work is a combined approach applied to MRI modalities implemented with AI strategies for autism classification using a synthesis of surface parameters, volumetric feature extraction using sMRI, functional newlineconnectivity features using rs-fMRI images, and IQ parameters.