Enhancing Autism Spectrum Disorder Detection and Classification with GAN-Driven Autoencoder Models

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder of highly variable symptoms, characterized by difficulties in communication, social interaction, and repetitive behaviors. Most current diagnostic approaches are based on behavioral assessments and clinical evaluations, which are rather subjective and may also require much time. Besides, traditional machine learning models fail to grasp the high-dimensional, intricate patterns of data in ASD, therefore seriously undermining the effectiveness in classification. This brings about an avid requirement for more efficient and reliable methodologies of ASD diagnosis. To tackle the above challenges, we propose here a novel method that exploits GAN s, combined with autoencoders (AE), for improved ASD detection and classification. The introduced GAN+AE model works in the feature extraction improvement direction—to learn complicated and high-level data representations. Compared to the conventional diagnostic features of ASD with added generation powers of GAN, our model presents improved performances in the classification of ASD cases to the best of their accuracy. Most significantly, compared with traditional machine-learning models, GAN+AE scored higher in terms of precision, recall, and F1-score; it decreases false negatives and maximizes ASD diagnostic accuracies—a promise for an earlier diagnosis tool for ASD. This article attempts to highlight the effectiveness of the deep learning models in overcoming the limitations of the traditional ASD detection approaches.