The role of large language models in recommending intervention disciplines and interventional therapies
10.3969/j.issn.1008-794X.2025.11.007
- VernacularTitle:大语言模型对介入科及介入治疗的推荐作用
- Author:
Zihan ZHAO
1
;
Zhihui CHANG
Author Information
1. 110004 辽宁沈阳 沈阳中国医科大学附属盛京医院放射科
- Keywords:
large language model;
artificial intelligence;
interventional therapy;
interventional radiology
- From:
Journal of Interventional Radiology
2025;34(11):1204-1209
- CountryChina
- Language:Chinese
-
Abstract:
Objective To explore the current status of big language models in recommending intervention disciplines and interventional therapies.Methods Four mainstream models,including Kimi k1.5,Doubao,DeepSeek-Rl and ChatGPT-4o were selected,and the standardized questions("recommended departments"and"treatment methods")for 18 common interventional diseases were designed.Through three repeated tests,the recommendation data were collected,and SPSS 25.0 was used to perform statistical analysis.Results Major language models recommended interventional physicians and their treatment methods to a certain extent,among which lower limb arterial occlusion and Stanford B-type aortic dissection in circulatory system diseases got the highest recommendation rates(100％)in interventional medicine and interventional treatment respectively,ranking among the top three recommendations.In contrast,benign prostatic hyperplasia had the lowest recommendation rate,with no recommendations for either interventional department or interventional treatment.Conclusion Large language models show disease-specific differences in recommendations for interventional radiology.Among the diseases involved in this study,circulatory system diseases have the highest referral degree,while urogenital system diseases have the lowest.The training data of the models need to further strengthen the coverage of scenario data related to the interventional department and its treatment methods,so as to gradually improve the social awareness of the interventional department.
