Investigating the diagnostic accuracy of GPT‐4's novel image analytics feature in dermatology.

The article explores the diagnostic accuracy of GPT-4's novel image analytics feature in dermatology, highlighting its potential in integrating textual and visual information for diagnosis. The study collected 100 clinical images from DermNet™ and found that GPT-4 correctly diagnosed 23% of images and included the correct diagnosis in 53% of cases. However, the tool displayed limited sensitivity for malignant and precancerous images, indicating its current clinical limitations. The study emphasizes the importance of caution in relying solely on AI platforms like GPT-4 for medical advice due to potential misdiagnoses and patient distress. [Extracted from the article]