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

## TL;DR

This study shows how AI can be used to create a decision-making tool for prostate cancer brachytherapy meetings, matching real-world decisions with 100% accuracy.

## Contribution

The first application of a custom AI tool using RAG for prostate cancer brachytherapy decisions in a multidisciplinary meeting.

## Key findings

- The AI tool matched the real-world MDM decisions for all 40 patients.
- The custom GPT was built using local guidelines and is publicly available on GitHub.
- The tool can streamline MDMs by preselecting optimal treatment strategies.

## 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.