Related Experiment Video
Updated: May 14, 2026
03:14
Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
Published on: December 6, 2024
The Role of Large Language Models in the Promotion of Minimally Invasive Interventional Radiologic Methods in
Iason Psilopatis1, Julius Emons2, Kleio Vrettou3
1Department of Gynecology and Obstetrics, University Hospital Basel, University of Basel, 4056 Basel, Switzerland.
Journal of Clinical Medicine
|May 13, 2026
Summary
Large language models (LLMs) can promote minimally invasive interventional radiology (IR) for gynecologic conditions, but their information quality varies. OpenEvidence and ChatGPT performed best, while Google Gemini showed weaker results.
Area of Science:
- Interventional Radiology
- Artificial Intelligence
- Gynecology and Obstetrics
Background:
- Minimally invasive interventional radiology (IR) provides uterus-preserving treatments for gynecologic conditions like fibroids, adenomyosis, and postpartum hemorrhage.
- Underutilization of these IR methods stems from limited clinician and patient awareness.
- Large language models (LLMs) present a potential solution for disseminating information on these treatments.
Purpose of the Study:
- To assess the effectiveness of current LLMs in addressing knowledge gaps and raising awareness of minimally invasive IR in gynecology and obstetrics.
- To evaluate LLM performance in providing accurate, complete, safe, and patient-centered information.
Main Methods:
- Three LLMs (OpenEvidence, ChatGPT, Google Gemini) were queried using a ten-question instrument.
- Responses were analyzed for accuracy, completeness, safety, and patient-centered communication.
Main Results:
- All LLMs accurately identified treatments for uterine fibroids, adenomyosis, and postpartum hemorrhage.
- OpenEvidence and ChatGPT provided more detailed and clinically nuanced information than Google Gemini.
- OpenEvidence and ChatGPT demonstrated superior accuracy, completeness, and safety, with Google Gemini showing inconsistent performance.
Conclusions:
- LLMs show potential for promoting minimally invasive IR in gynecology and obstetrics, but output quality varies significantly.
- Further refinement and integration of evidence-based sources are crucial for clinical application.
- Collaboration between AI developers and medical professionals is essential to optimize LLM utility in this field.