Large language model (ChatGPT) as a support tool for breast tumor board

Area of Science:

• Oncology
• Artificial Intelligence
• Medical Informatics

Background:

• Large language models (LLMs) are gaining attention in both public and scientific spheres.
• The application of LLMs in clinical decision-making, particularly in oncology tumor boards, is an emerging area of research.

Purpose of the Study:

• To evaluate the efficacy of ChatGPT (specifically GPT-3.5) as a decision support tool for breast cancer tumor boards.
• To compare ChatGPT's management recommendations against final decisions made by a multidisciplinary tumor board.

Main Methods:

• Clinical data from ten breast cancer patients were input into ChatGPT-3.5.
• ChatGPT was prompted to recommend patient management.
• Recommendations were compared to tumor board decisions and independently graded by two senior radiologists on summarization, recommendation, and explanation (1-5 scale).

Main Results:

• ChatGPT's recommendations matched tumor board decisions in 70% of cases.
• Radiologists' mean scores for ChatGPT's summarization, recommendation, and explanation ranged from 3.7 to 4.6.
• The study included patients with invasive ductal carcinoma, ductal carcinoma in-situ, and phyllodes tumor.

Conclusions:

• This proof-of-concept study demonstrates the potential of LLMs as decision support tools in breast tumor boards.
• Clinicians should become familiar with the benefits and risks associated with using LLMs in clinical practice.
• Further research is warranted to explore the integration and validation of LLMs in oncological decision-making processes.