AI-Assisted Autism Diagnosis

In March, 2025, researchers leveraged a large language model (LLM) to identify the most relevant behavioural patterns for autism diagnosis, refining traditional clinical assessment methods.

Key Points

AI in Diagnosis: A transformer-based LLM was fine-tuned on over 4,000 clinician reports, highlighting key behavioral indicators without predefined diagnostic outcomes.
Most Relevant Behaviors: The model identified repetitive behaviors, special interests, and perception-based behaviors as the strongest predictors of autism, rather than the social deficits emphasized in DSM-5 guidelines.
Enhanced Diagnostic Tools: The LLM framework aims to support clinicians by providing more objective diagnostic insights aligned with empirical clinical observations.
Broader Impact: This approach could help refine diagnostic methodologies for psychiatric, mental health, and neurodevelopmental disorders, where clinical judgment is a key factor.