GPT-4 and medical image analysis: strengths, weaknesses and future directions

ChatGPT (Generative Pre-trained Transformer) is an artificial intelligence (AI) language model developed by OpenAI. GPT-4 is the newest version of ChatGPT released on March 14, 2023 and has been reported to have a broader knowledge base as well as improved problem-solving ability. GPT-4 has also been reported to be less easy to fool, and is capable of processing 8 times more words. The usages for ChatGPT continue to grow, and new applications of the langue learning model continue to be found. Due to the black box nature of AI models, interpretation of GPT-4 outputs must be made with caution to ensure that no errors have been made. Particularly in healthcare delivery and medicine, where policies and procedures are frequently revised, GPT-4 algorithms comments may be out-of-date or incorrect. Out of the new features introduced in GPT-4, the most important feature may be its new ability to analyze images. This could potentially help doctors to diagnose and treat patients quickly and accurately, especially in areas where access to medical professionals may be limited. To examine GPT-4’s image diagnostic ability, we provided it with a variety of common medical imaging modalities: from chest X-rays, magnetic resonance images (MRI), to optical coherence tomography (OCT) images. All in all, although significant advancements and further research is still required, the future of automated medical image analysis is highly promising.