GAN-Enhanced Multimodal Artificial Intelligence for Autism Spectrum Disorder Screening: A Comprehensive Survey

Abstract Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by difficulties in social communication and interaction, together with restricted or repetitive patterns of behavior. Early screening is important for facilitating timely clinical assessment and intervention; however, conventional screening approaches often depend on clinical expertise, behavioral observation, developmental history, and standardized assessment tools. Recent advances in artificial intelligence (AI), machine learning (ML), deep learning (DL), and Generative Adversarial Networks (GANs) have created new opportunities for data-driven and automated ASD screening. This survey provides a comprehensive review of AI-based approaches for ASD screening, with particular emphasis on multimodal learning and the integration of heterogeneous data sources. Behavioral, genetic, environmental, neuroimaging, physiological, and clinical data modalities are examined, along with approaches for combining information across multiple sources. The survey further reviews the role of GANs and related generative approaches in synthetic-data generation and data augmentation, particularly in addressing challenges associated with limited and imbalanced datasets. Multimodal fusion strategies, multi-input neural networks, and emerging deep learning approaches are also discussed in the context of ASD screening. The reviewed literature indicates that multimodal AI has the potential to provide a more comprehensive representation of ASD-related characteristics than single-modality approaches. However, important challenges remain, including limited and heterogeneous datasets, class imbalance, multimodal data integration, GAN training instability, synthetic-data quality, privacy and security, model interpretability, generalizability, and limited external clinical validation. This survey identifies these challenges and highlights research directions toward reliable, explainable, privacy-preserving, and clinically validated AI-based multimodal ASD screening systems. Keywords: Autism Spectrum Disorder (ASD); Artificial Intelligence; Machine Learning; Deep Learning; Generative Adversarial Networks (GANs); Multimodal Learning; ASD Screening; Data Augmentation; Synthetic Data