Construction and Application of a Multicenter Traditional Chinese Medicine Proctology Disease Data Platform Based on Multimodal Large Models
10.13288/j.11-2166/r.2026.07.012
- VernacularTitle:基于多模态大模型的多中心中医肛肠优势病种专病数据平台的构建与应用
- Author:
Yuxin ZHU
1
;
Liping ZHAO
2
;
Jiafa LU
3
;
Huiting ZHU
1
;
Xia YANG
1
;
Lei DU
1
;
Kang DING
1
Author Information
1. Nanjing Hospital of Chinese Medicine affiliated to Nanjing University of Chinese Medicine，Nanjing，210022
2. Traditional Chinese Medicine Administration，Jiangsu Commission of Health
3. Jiangsu Provincial Health Information Center
- Publication Type:Journal Article
- Keywords:
hemorrhoids;
multimodal large model;
specialized disease dataset;
dominant diseases of TCM;
real-world study
- From:
Journal of Traditional Chinese Medicine
2026;67(7):770-775
- CountryChina
- Language:Chinese
-
Abstract:
This paper has constructed a traditional Chinese medicine （TCM） specialized disease dataset platform for mixed hemorrhoids based on a multimodal large model， and the preliminary application has been validated. The platform uses StarRocks to establish a four-level data warehouse system， enabling the aggregation， cleaning， and standardization of multi-source heterogeneous data. Using DeepSeek-R1-Distill-Qwen-7B as the base model， domain fine-tuning is performed through low-rank adaptation （LoRA） technology. Combined with LLaMA-3.3 natural language processing and reasoning chain techniques， the platform enables intelligent parsing and structured extraction of unstructured TCM medical records. It accurately identifies six major categories and 28 subcategories of entities， including symptoms and syndromes， with a fine-tuned model F1 score of 93.8%. The platform has established a high-quality specialized disease dataset containing more than 50,000 medical records and has been applied in a real-world study involving 17,831 patients， preliminarily verifying the efficacy of TCM heritage surgery.
