Evaluating the Accuracy and Diagnostic Reasoning of Multimodal Large Language Models in Interpreting Neuroradiology Cases From RadioGraphics

Key Points

• The research aims to assess how accurately multimodal large language models interpret neuroradiology cases compared to experts.
• Compared performance of multimodal LLMs and neuroradiologists in interpreting neuroradiology cases.
• Analyzed reasoning for differential diagnoses provided by LLMs.
• Evaluated impact of absence of clinical history on LLM performance.
• Multimodal LLMs significantly underperformed compared to neuroradiologists.
• LLMs exhibited unsatisfactory reasoning for differential diagnoses.
• Performance declined further for cases lacking textual input of clinical history.