Generative Models in Early Detection of Neurodevelopmental

Disabilities: A Comprehensive Review of Applications,

Innovations, and Challenges

Ika Hesti Agustin1,* Dwi Agustin Retnowardani 2

1 Department of Mathematics, University of Jember, Jember, East Java, Indonesia 2 Universitas PGRI Argopuro Jember

Received: December 10, 2025 Revised: January 27, 2026 Accepted: March 08, 2026 ⋆ Corresponding author

ABSTRACT

Neurodevelopmental disorders are a broad category that estimates fifteen million people and include autism spectrum

disorder (ASD), attention-deficit/hyperactivity disorder (ADHD), and intellectual disabilities that, if not found at

an early age, present substantial lifelong challenges. Modern technologies in artificial intelligence with generative

models mean new possibilities in early diagnostics and prevention. This review aims to review the biomarker

potentials of generative models, including GANs, VAEs, and diffusion models, in the early diagnosis of neurodevelopmental

disabilities. Having synthesized what is currently known about these models, we explore how the models

improve diagnostic precision, minimize the use of invasive procedures, and manage data deficiency. The significant

applications discussed involve generative models in analyzing neuroimaging data, modeling speech and behavior,

and synthesizing new datasets that are valuable in handling privacy issues and biased datasets. In addition, this paper

discusses some of the limitations associated with generative model deployment in clinical practice; these include

interpretability, model stability, and the fact that the models rely on extensive and diverse datasets. Finally, we bridge

the gap by looking into the future and discussing what future research could bring and ethical concerns regarding

generative models and their potential to revolutionize handling cases of early neurodevelopmental disorders and

enable early, more effective interventional approaches.

Keywords: Neurodevelopmental disorders Generative models Early diagnosis Artificial intelligence Personalized

medicine

1. INTRODUCTION

Neurodevelopmental disorders (NDDs) are a broad group

of developmental conditions that influence the thinking and

learning abilities, behavior, and emotions of an individual.

These disorders include autism spectrum disorder (ASD),

attention-deficit/hyperactivity disorder (ADHD), and intellectual

disability (ID). They are estimated to affect a substantial

number of individuals worldwide and can present long-term

challenges, particularly when early diagnosis and treatment

are unavailable. Therefore, early diagnosis and timely intervention

are essential because they can improve treatment

outcomes and support better developmental trajectories. In

this context, the development of new diagnostic approaches

has become an important research direction.

The development of artificial intelligence (AI) technologies

has advanced rapidly and has enabled major progress in

healthcare. Among these technologies, generative models, including

generative adversarial networks (GANs), variational

autoencoders (VAEs), and diffusion models, are especially