Figure 4. This figure displays the overall diagnostic accuracy of all AI models (ChatGPT 3.5, ChatGPT 4.0, and Google Gemini) across different neuroradiology subgroups. The combined accuracy for brain-related questions was 62%. In the head and neck category, accuracy improved to 69.14%. The accuracy was 52.03% for spine-related questions and 48.02% for non-interpretive skills.