Explainable AI for ASD Detection using fMRI Data

Project Details

• Student(s): Ali Assi
• Advisor(s): Dr. Lina Abou Abbas
• Department: Electrical & Computer
• Academic Year(s): 2024-2025

Abstract

Autism spectrum disorder (ASD) is a complex neurodevelopmental condition characterized by per sistent challenges in social interaction, communication, and restricted or repetitive behaviors. Early and accurate diagnosis is crucial for providing timely interventions and improving outcomes. Ma chine learning (ML) methods, particularly deep learning approaches, have shown promising potential for automated detection of ASD through analysis of neuroimaging data. The aim of this study is to explore state-of-the-art ML techniques combined with Explainable Artificial Intelligence (XAI) methods for the detection and interpretability of ASD diagnosis based on functional magnetic res onance imaging (fMRI) data. This research emphasizes the significance of model interpretability, aiming to bridge the gap between clinical applicability and the technical complexity of deep learning models. Findings suggest that the integration of XAI techniques with ML models not only achieves reliable diagnostic accuracy but also provides valuable insights into neurological patterns associated with ASD. Furthermore, enhancing model transparency through explainability significantly increases clinicians’ trust in the diagnostic outcomes. Future large-scale studies incorporating explainability techniques are essential to validate these findings and accelerate their translation into clinical prac tice, thus facilitating early, transparent, and effective ASD diagnosis.