Data - Driven Approaches for Detecting Autism Spectrum Disorder : A Comprehensive Review

Abstract

Autism Spectrum Disorder (ASD) is a neuro developmental ailment that affects significant portion of the global population, with symptoms including cognitive and behavioural impairments. Early and precise detection is critical for effective intervention and helps to minimize the overall impact of it, however, the traditional methods of ASD detection are time-consuming and dependent on clinical expertise. In recent years, Data Mining combined with Machine Learning (ML) and Deep Learning (DL) techniques have shown huge potential in refining the accuracy, speed, and efficiency of ASD diagnosis. This paper reviews the application of these advanced technologies, particularly the use of neuroimaging data such as fMRI, EEG, and MRI, in diagnosing ASD. By leveraging the automated feature extraction technique of these methods, predictive algorithms, and end-to-end learning, ML and DL models have demonstrated the potential to identify complex patterns in brain activity, offering insights into potential biomarkers for ASD. Several approaches have been presented to augment the detection accuracy where conventional machine learning methods have focused on data mining approaches along with image data. However, the accuracy of these systems for imaging data is affected due to complex brain patterns therefore; deep learning based methods have gained popularity because of their robust pattern learning methods. Thus, the exploitation of ML and DL with neuroimaging tools has the potential to enhance early diagnosis, paving way for timely interventions and enhanced results for patients with ASD. This review also highlights the challenges and future prospects of applying these technologies to ASD diagnosis, plus the need for large-scale datasets and the development of universally applicable frameworks.