Exploiting large language models for diagnosing autism associated language disorders and identifying distinct...

Area of Science:

• Computational linguistics
• Developmental psychology
• Artificial intelligence in healthcare

Background:

• Diagnosing language disorders in autism spectrum disorder (ASD) is challenging due to subjective traditional methods.
• Existing assessment tools for ASD-related language impairments often lack speed and precision.
• Variability in traditional assessments can hinder accurate and timely diagnosis of autism.

Purpose of the Study:

• To explore the use of large language models (LLMs) for enhancing the diagnosis of language disorders associated with autism.
• To improve the sensitivity and precision of autism diagnosis by leveraging natural language understanding (NLU).
• To identify and profile key linguistic features indicative of autism spectrum disorder.

Main Methods:

• Utilized large language models (LLMs) with natural language understanding (NLU) capabilities.
• Employed a zero-shot learning configuration to assess model performance without prior specific training on the dataset.
• Analyzed linguistic features in language samples to identify patterns associated with autism.

Main Results:

• The LLM-based method showed over a 10% increase in sensitivity and positive predictive value compared to baseline models.
• The framework demonstrated improved accuracy and applicability as a supplementary diagnostic tool.
• Identified ten key linguistic features, including echolalia and pronoun reversal, crucial for ASD diagnosis.

Conclusions:

• LLMs offer a promising approach to overcome limitations in diagnosing autism-related language disorders.
• The developed framework can serve as a valuable supplementary tool for clinicians.
• Identifying specific language features aids in a more precise diagnosis and personalized treatment for ASD.