SIR 2024
General IR
Chloe Cross, MD
Resident
Icahn School of Medicine At Mount Sinai
Financial relationships: Full list of relationships is listed on the CME information page.
Hayden Hofmann, BS
Medical Student
Keck School of Medicine of USC
Disclosure information not submitted.
Robert A. Lookstein, MD
Executive Vice Chair; Diagnostic, Molecular, and Interventional Radiology
Mount Sinai Hospital
Financial relationships: Full list of relationships is listed on the CME information page.
Kirema Garcia-Reyes, MD
Assistant Professor
Mount Sinai
Financial relationships: Full list of relationships is listed on the CME information page.
Aaron M. Fischman, MD, FSIR, FCIRSE, FSVM
Professor, Diagnostic, Molecular and Interventional Radiology
Icahn School of Medicine at Mount Sinai
Disclosure information not submitted.
Edward Kim, MD (he/him/his)
Professor of Radiology
Mount Sinai Health System
Financial relationships: Full list of relationships is listed on the CME information page.
Vivian Bishay, MD
IR
Mount Sinai Hospital System
Financial relationships: Full list of relationships is listed on the CME information page.
Rahul S. Patel, MD
Assistant Professor, Diagnostic, Molecular and Interventional Radiology
Mount Sinai Medical Center\n
Financial relationships: Full list of relationships is listed on the CME information page.
Dan Shilo, MD
Assistant Professor, Diagnostic, Molecular and Interventional Radiology
Mount Sinai Hospital
Disclosure information not submitted.
F Scott Nowakowski, MD
Professor of Radiology and Surgery
Mount Sinai Medical System
Disclosure information not submitted.
Rajesh I. Patel, MD
Assistant Professor, Diagnostic, Molecular and Interventional Radiology
Mount Sinai Hospital
Disclosure information not submitted.
Jenanan Vairavamurthy, MD
Assistant Professor, Interventional Radiologist
Keck School of Medicine, Univeristy of Southern California
Disclosure information not submitted.
Large language models like ChatGPT are being used by patients to understand disease processes and determine who to seek for treatment. Determining how these models portray Interventional Radiology (IR) to patients can give insight into the public perception of IR. The purpose of this study was to evaluate if ChatGPT suggests treatment by IR for common IR treated disease processes.
Materials and Methods:
A list of disease processes in various systems commonly treated by IR was selected from literature review and agreed upon by the authors. A standard prompt was developed to reflect a realistic patient question: “I have (disease process). What types of doctors can treat this?” Due to the probabilistic nature of the model, 3 repeat prompts to ChatGPT were completed for each disease process. IR suggestion, as well as in what order IR was suggested to the other specialties (rank) was noted. A one-way ANOVA test was performed to calculate variance amongst disease processes for both outputs.
Results:
A total of 69 prompts were performed (23 disease processes, 3 repeat prompts each). ChatGPT generated summaries suggested IR for treatment of the disease in 73.9% of outputs (51/69). On average IR was the third suggested specialty (rank = 3.3). One-way ANOVA revealed that whether IR was suggested significantly varied for each disease process (F=2.5, p=0.003). However, there was no significant variance for how IR ranked compared to other specialties for each disease process (F=1.9, p=0.052). Interestingly, IR was the first ranked output 3/3 times for splenic artery aneurysm treatment but not mentioned at all for the treatment of pulmonary embolism or bone masses. (Table 1)
Conclusion:
Large language models, specifically ChatGPT, acknowledge IR’s role in the treatment of several but not all the disease processes we are known to treat. It is important for interventional radiologists to understand how large language models may aid in the public perception of IR.