Accuracy of ChatGPT in Neurolocalization

Area of Science:

• Neurology
• Artificial Intelligence
• Medical Education

Background:

• Artificial intelligence (AI) chatbots, such as ChatGPT, possess advanced communication skills and extensive knowledge bases.
• The application of AI in medicine, particularly in neurolocalization, requires robust clinical reasoning alongside neuroanatomical expertise.

Purpose of the Study:

• To evaluate the capabilities of ChatGPT in performing neurolocalization tasks.
• To assess the accuracy of ChatGPT's responses in clinical case scenarios.

Main Methods:

• Forty-six text-based neurolocalization case scenarios were presented to ChatGPT-3.5.
• Seven neurosurgeons evaluated the accuracy of ChatGPT's responses using a 5-point scoring system.

Main Results:

• ChatGPT-3.5 demonstrated an accuracy of 84.8% in providing "completely correct" or "mostly correct" answers.
• Analysis indicated consistent scoring among the neurosurgical evaluators.

Conclusions:

• The current accuracy of AI chatbots in neurolocalization is promising but insufficient for direct clinical application in patient care.
• Concise and precise interactions with AI tools are recommended to enhance response accuracy.
• AI holds potential for future breakthroughs in medical applications.