ChatGPT-3.5, a large language model, has been tested as a supporting tool for physicians to manage breast cancer patients. The study included ten women recently diagnosed with breast cancer and presented at the breast tumor board at a medical institution. Relevant clinical data and imaging results were collected, and clinical vignettes were generated summarizing the patient’s condition. The vignettes were then inserted into ChatGPT-3.5 and used to recommend the next most appropriate step in management. The chatbot’s recommendations were compared to the tumor board’s decisions, and responses were graded based on summarization, clinical recommendation, and explanation. The study demonstrates the feasibility of using large language models to support patient care tasks. However, further studies are required to establish the actual performance of LLMs and their strengths and pitfalls. It is also essential for clinicians to be familiar with the pros and cons of this technology to smoothly integrate its use into clinical scenarios.
The article mentions ChatGPT-3.5, a large language model designed by OpenAI, a company founded by entrepreneurs such as Elon Musk and Sam Altman. The model’s purpose is to assist in natural language processing and can be used for tasks such as language translation, summarization of long texts, and question-answering. This research demonstrated ChatGPT-3.5’s potential as a support tool for physicians in managing breast cancer patients.
The study’s authors, Y.B., V.S., and M.S.L., reviewed the chatbot’s performance and compared its responses to the tumor board’s decisions. M.S.L. is a senior breast radiologist with experience in tumor boards and evaluated the chatbot’s clinical recommendations. Ey.K., another senior radiologist, graded the responses independently to avoid bias introduction. The study’s grading system was based on three distinct categories: summarization, clinical recommendation, and explanation. This allowed the authors to analyze the chatbot’s performance more accurately.
