EASDM: Explainable Autism Spectrum Disorder Model Based on Deep Learning

• Published in: Journal of Disability Research Vol. 3; no. 1
• Main Authors: Atlam, El-Sayed, Masud, Mehedi, Rokaya, Mahmoud, Meshref, Hossam, Gad, Ibrahim, Almars, Abdulqader M.
• Format: Journal Article
• Language: English
• Published: 03.02.2024
• DOI: 10.57197/JDR-2024-0003
• ISSN: 1658-9912

A neuro-developmental disorder known as autism spectrum disorder (ASD) affects a significant portion of the global population. Those with ASD frequently struggle to interact and communicate with others and may engage in restricted or repetitive behaviors or interests. The symptoms of autism begin early in childhood and can continue into adulthood. Machine learning and deep learning (DL) models are employed in clinical research for the early identification and diagnosis of ASD. However, the majority of the existing models lack interpretability in their results for ASD diagnosis. The explainable artificial intelligence (XAI) concepts can be used to provide transparent and understandable explanations for models’ decisions. In this work, we present an explainable autism spectrum disorder model based on DL for autism disorder detection in toddlers and children. The primary objective of this study is to better understand and interpret the classification process and to discern the significant features that contribute to the prediction of ASD. The proposed model is divided into two distinct components. The first component employs a DL model for autism disorder detection. The second uses an XAI technique known as shapley additive explanations (SHAP) to emphasis key characteristics and explain the model’s outcomes. The model showed perfect performance on the training set, with an accuracy of 1 and a receiver operating characteristic score of 1. On the test set, the model achieved an accuracy score of 0.9886, indicating that it performed nearly as well as on the training set. The experimental results demonstrate that the proposed model has the capability to accurately predict and diagnose ASD while also providing explanatory insights into the obtained results. Furthermore, the results indicate that the proposed model performs competitively compared to the state-of-the-art models in terms of accuracy and F1-score. The results highlight the efficacy and potential of the proposed model in accurately predicting ASD in binary classification tasks.