Hidden flaws behind expert-level accuracy of multimodal GPT-4 vision in medicine
This highlights critical flaws in AI reasoning for clinical use, emphasizing the need for deeper evaluation before deployment in medicine.
The study analyzed GPT-4V's performance on medical image challenges, finding it matches human physicians in multi-choice accuracy (81.6% vs. 77.8%) but often provides flawed rationales for correct answers, with 35.5% of correct choices having errors, particularly in image comprehension (27.2%).
Recent studies indicate that Generative Pre-trained Transformer 4 with Vision (GPT-4V) outperforms human physicians in medical challenge tasks. However, these evaluations primarily focused on the accuracy of multi-choice questions alone. Our study extends the current scope by conducting a comprehensive analysis of GPT-4V's rationales of image comprehension, recall of medical knowledge, and step-by-step multimodal reasoning when solving New England Journal of Medicine (NEJM) Image Challenges - an imaging quiz designed to test the knowledge and diagnostic capabilities of medical professionals. Evaluation results confirmed that GPT-4V performs comparatively to human physicians regarding multi-choice accuracy (81.6% vs. 77.8%). GPT-4V also performs well in cases where physicians incorrectly answer, with over 78% accuracy. However, we discovered that GPT-4V frequently presents flawed rationales in cases where it makes the correct final choices (35.5%), most prominent in image comprehension (27.2%). Regardless of GPT-4V's high accuracy in multi-choice questions, our findings emphasize the necessity for further in-depth evaluations of its rationales before integrating such multimodal AI models into clinical workflows.