ChatGPT and Interventional Radiology

Artificial intelligence (AI) comprises software algorithms that can complete complex tasks that usually require human capability for their execution. ChatGPT is a large language model (LLM) developed by OpenAI and released for use in November 2022. The chat box uses a generative pretrained transformer (GPT), which generates new text using a transformer model of deep learning trained on existing human-generated text. The output depends on the number of tunable parameters and the training data fed into the model.

ChatGPT in interventional radiology (IR) is used for patient and physician communication.[1] It can introduce patients to IR providers and provide information about the different minimally invasive therapeutic interventions in a more readable and understandable form without language limitations. It can be used for patient scheduling, consenting, and postprocedure monitoring, including screening for postprocedure complications.

For physicians, ChatGPT can simplify clinical documentation, help in the recruitment of patients for clinical trials, and also be used as a medical decision support tool. ChatGPT can facilitate evidence-based clinical decision-making by evaluating the existing literature on the subject. This reduces the time required for an extensive literature search in a busy IR facility. On the research front, it can help in searching the literature, making tables and graphs, and selecting the optimum statistical analysis. ChatGPT can also be used to provide suggestions for the procedure, such as the access approach and the preferred hardware for complex IR procedures.

The advantage of ChatGPT is that it can perform various tasks in several languages at different levels of health literacy without any significant time lag. It can summarize a large volume of information and has interactive capability. The limitations of ChatGPT are that inaccurate results or “hallucinations” may be generated due to the inability to infer the meaning or the context of the query and the lack of the required data. The potential threat of this new software is the inability to maintain patient privacy and ethical standards, misinformation, and biases due to inadequate training data. Corrective measures including human review, fine-tuning, and prompt engineering can adjust the model's inherent knowledge and information.[2] [3]

ChatGPT should be used in a judicious and collaborative manner with human inputs to provide specialized patient care. There is a definite need for human oversight to ensure the credibility and reliability of this exciting new tool.

We conclude with a humorous limerick about an interventional radiologist generated from ChatGPT:

“An interventional radiologist with great cheer,

Could fix any blockage-no fear,

With a poke and a thread,

He rewired instead,

And the surgeons all gave him a cheer!”

Publikationsverlauf

Artikel online veröffentlicht:
01. April 2025

• References

• 1 Campbell IV WA, Chick JFB, Shin D, Makary MS. Understanding ChatGPT for evidence-based utilization in interventional radiology. Clin Imaging 2024; 108: 110098
• 2 Javan R, Kim T, Mostaghni N, Sarin S. ChatGPT's potential role in interventional radiology. Cardiovasc Intervent Radiol 2023; 46 (06) 821-822
• 3 Zaki HA, Mai M, Abdel-Megid H. et al. Using ChatGPT to improve readability of interventional radiology procedure descriptions. Cardiovasc Intervent Radiol 2024; 47 (08) 1134-1141