The number of people with conditions on the autistic spectrum is rapidly increasing. Unfortunately, they may find it difficult to obtain early diagnosis and intervention to prevent problematic behaviors, which can lead to social isolation and financial strain for the entire family. The global rise in autism spectrum disorder (ASD) diagnoses has created a greater need for new and effective diagnostic methods. Technological advancements have prompted healthcare practitioners to investigate the use of artificial intelligence (AI) for diagnosing ASD. AI can enable accurate, rapid diagnosis by analyzing detailed data on developmental milestones, behavioral traits, and medical histories. In this study, a framework is developed to support the diagnosis of ASD using machine learning and deep learning models. The framework was tested with a standard dataset from the North Cairo Governorate, Egypt, collected across multiple institutions. The dataset includes symptomatic data gathered using the modified checklist for autism in toddlers (M-CHAT-R), focusing on selected characteristics starting with “Q” and the “class” label. The features were standardized using min-max normalization for each sequence dimension. The SHapley Additive exPlanations method was used to identify key features that improve classification accuracy. These features were processed using eXtreme Gradient Boosting, random forest, and the proposed models, including convolutional neural networks (CNNs) and long short-term memory (LSTM) models, for early ASD diagnosis. The CNN–LSTM model achieved 95% accuracy in detecting ASD. The system’s framework demonstrates significant potential to support ASD diagnosis.
Al-Nefaie et al. (Thu,) studied this question.