Harnessing Artificial Intelligence to Code a Decision-Making Aid for the Prostate Cancer Brachytherapy Multidisciplinary Meeting Using Retrieval-Augmented Generation

Abstract

Introduction We have previously published the first application of artificial intelligence (AI) to streamline the real-world multidisciplinary meeting (MDM) process, finding that ChatGPT (OpenAI) could accurately recommend a treatment option in perfect concordance with the European Association of Urology (EAU) guidelines. ChatGPT now allows users to formally create their own generative pretrained transformer (GPT) tailored to a specific role using retrieval-augmented generation (RAG). In this paper, we seek to utilize this technology for the first time in the prostate cancer multidisciplinary team (MDT) setting, to create a custom decision-making aid GPT for prostate cancer patients being treated with brachytherapy. Methods ChatGPT 4.0 was prompted to create a decision-making custom GPT using RAG to select brachytherapy treatment options for prostate cancer patients. While international guidelines exist, we opted to use our well-established local Royal Surrey guidelines as prior knowledge to the custom GPT. Results The custom GPT created is available on GitHub. Forty patients were discussed, with the AI-generated tool suggesting the same brachytherapy treatment strategy as that of the real-world MDM in all (40/40) cases. Conclusion We believe that this simple study has demonstrated a very practical application of this novel AI technology. This has the capacity to reinvent the paradigm in the preparation of patient cases prior to a cancer board or MDM, as cases can be preselected for the optimal treatment strategy with 100% accuracy prior to discussion by clinicians in the MDM, streamlining the process and facilitating a nuanced discussion by eliminating irrelevant treatment strategies beforehand.

Ähnliche Arbeiten