# Early autism detection: a review of emerging technologies, biomarkers, and explainable AI approaches

## TL;DR

This paper reviews emerging technologies and biomarkers for early autism detection, emphasizing the potential of eye gaze analysis and explainable AI.

## Contribution

The paper highlights eye gaze analysis combined with explainable AI as a novel, under-explored approach for early ASD detection.

## Key findings

- Eye gaze analysis offers a cost-effective, non-invasive method for early ASD detection.
- Explainable AI can improve transparency and clinical confidence in autism diagnostics.
- Current research gaps include limited interpretability in neuroimaging models and ethical issues in genomics.

## Abstract

Autism Spectrum Disorder (ASD) presents as a complicated neurodevelopmental disorder which leads to social communication challenges and repetitive behavioral patterns. Early identification of ASD is crucial to facilitate early intervention that can make a large positive impact on long-term developmental outcomes. With the advent of artificial intelligence (AI) and data-driven diagnoses, there is increased interest in combining machine learning methods with biological and behavioral signatures to detect early ASD. This review provides an overview of broad classes of biomarkers—behavioral, neuroimaging, genetic, and eye gaze—and their respective methodologies, clinical applications, and diagnostic accuracy. For each of these biomarker domains, the research gap has been identified as existing for instance limited interpretability in neuroimaging models, genomics-related ethical and data accessibility issues, and innovation saturation for behavioral measurement. A comparative analysis highlights eye gaze analysis as a promising but under-explored option, providing a balance of cost-effectiveness, non-invasiveness, and potential for real-time, objective measurement. In addition, the application of Explainable AI (XAI) methodologies across these biomarker fields is discussed in order to meet the pressing need for transparency, clinical confidence, and decision-making support. This review makes a final call for further exploration of eye gaze-based models enriched by XAI methods as a future research direction towards filling the gap between algorithmic innovation and real-world, interpretable diagnostics in the context of ASD research.