Evaluation of large language models in clinical neuroanatomy: a comparative scoring analysis based on accuracy, concordance, insight, and anatomical terminology accuracy

Key Points

• The research compares the performance of large language models in neuroanatomical diagnostics, focusing on accuracy and terminology use.
• Utilized structured scoring frameworks like ACI and ATA for evaluation.
• Analyzed diagnostic performance based on accuracy and anatomical terminology.
• Compared ChatGPT-4 and Gemini 2.5 in various neuroanatomical scenarios.
• ChatGPT-4 demonstrated strong and stable diagnostic performance with high accuracy.
• Gemini 2.5 showed inconsistent results, particularly sensitive to prompt changes.
• Both models can be valuable in clinical and educational settings.