Artificial intelligence-driven diagnosis of autism spectrum disorder in children

Autism Spectrum Disorder (ASD) significantly impacts society by highlighting the need for inclusive education, healthcare, and employment systems that support neurodiversity. This challenges societal norms and promotes greater awareness, understanding, and acceptance, encouraging communities to become more inclusive and supportive of individuals with diverse abilities.

The main novelty is to identify and explore the key factors affecting ASD in Saudi Arabia and Egypt, aiming to improve early diagnosis through Explainable Artificial Intelligence (XAI) techniques, specifically SHapley Additive exPlanations (SHAP), LIME (Local Interpretable Model-agnostic Explanations), and Permutation Feature Importance (PFI). The research primarily employed decision tree (DT) and K-Nearest Neighbors (KNN) models combined with explainable AI (XAI), leading to significant improvements in diagnostic accuracy. The system was applied to a real dataset collected from various locations across Saudi Arabia, which is publicly available on Kaggle. Additionally, another ASD dataset from the Data Science Bank repository, sourced from participants in North Cairo Governorate, Egypt, was used for testing the system. Before analysis, the data was validated by removing outliers, filling missing data, and confirming the relevance of selected features. The study aims to enhance early diagnosis through XAI methods, including SHAP, LIME, and PFI.

The results show that KNN, when combined with XAI, achieved a high accuracy of 97% on the Saudi Arabia dataset and 92% on the Egypt dataset.

This approach has proven to be effective in developing more accurate and straightforward AI models for ASD diagnosis. It demonstrates that integrating advanced AI techniques with practical clinical applications can significantly improve the healthcare system in Saudi Arabia, leading to earlier ASD detection, better-informed treatment plans, and ultimately, an improved quality of life for those affected.