عبدالسلام السوداني

Earlier predictions of autism spectrum disorder (ASD) status in children are critical from both clinical an‎d societal perspectives, as late diagnosis can attenuate the efficacy of early intervention attempts while preventing developmental support later in life. While machine learning applications are gaining momentum in the field of autism research in general, most of these findings originate from heuristic models that simplify behavioral data into isolated an‎d independent features, losing clinical interpretability an‎d being unable to account for the developmental character of children’s behavior. This is an important limitation, as symptoms of autism often present in a clustered/clustered pattern an‎d not individually. To address this challenge, the current study re-conceptualizes the diagnostic problem by representing childrenʹs behavioral symptoms as structured an‎d interconnected paths that retain their developmental associations. The study is based on an artificial intelligence model including a Ran‎dom Forest classifier combined with pathway-driven behavior profiling an‎d similarity-based interpretation. As such, the study demonstrates how artificial intelligence can be used more transparently, robustly, an‎d clinically relevant to aid early autism prediction in children an‎d to improve its clinical decision support in real settings.

Earlier predictions of autism spectrum disorder (ASD) status in children are critical from both clinical an‎d societal perspectives, as late diagnosis can attenuate the efficacy of early intervention attempts while preventing developmental support later in life. While machine learning applications are gaining momentum in the field of autism research in general, most of these findings originate from heuristic models that simplify behavioral data into isolated an‎d independent features, losing clinical interpretability an‎d being unable to account for the developmental character of children’s behavior. This is an important limitation, as symptoms of autism often present in a clustered/clustered pattern an‎d not individually. To address this challenge, the current study re-conceptualizes the diagnostic problem by representing childrenʹs behavioral symptoms as structured an‎d interconnected paths that retain their developmental associations. The study is based on an artificial intelligence model including a Ran‎dom Forest classifier combined with pathway-driven behavior profiling an‎d similarity-based interpretation. As such, the study demonstrates how artificial intelligence can be used more transparently, robustly, an‎d clinically relevant to aid early autism prediction in children an‎d to improve its clinical decision support in real settings. , پيش‌بيني زودهنگام , هوش مصنوعي , يادگيري ماشين , مدل‌سازي مسير رفتاري , جنگل تصادفي

Early Prediction , Artificial Intelligence , Behavioral Pathway Modeling , Machine Learning , Autism Spectrum Disorder

Abdulsalam Alsoodani

Hassan Naderi