慢性心衰预测模型构建方法、装置、设备及存储介质
By training a biomedical large language model and performing instruction fine-tuning and low-rank adaptive fine-tuning, the problems of overfitting and poor prediction accuracy in chronic heart failure prediction models were solved, achieving high accuracy and robust prediction with limited data.
CN120809256BActive Publication Date: 2026-07-17GUANGZHOU UNIVERSITY OF CHINESE MEDICINE
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- GUANGZHOU UNIVERSITY OF CHINESE MEDICINE
- Filing Date
- 2025-07-02
- Publication Date
- 2026-07-17
AI Technical Summary
Technical Problem
Existing chronic heart failure prediction models suffer from problems such as overfitting and poor prediction accuracy due to their complex model architecture and large number of parameters, making it difficult to collect sufficient high-quality clinical data.
Method used
By acquiring biomedical literature sample data, multi-disease clinical sample data, and target disease clinical sample data, a biomedical large language model is trained and generated. Then, instruction fine-tuning and low-rank adaptive fine-tuning are performed to construct a chronic heart failure prediction model.
Benefits of technology
With limited real-world clinical data samples, the model's predictive accuracy, stability, and robustness were improved.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN120809256B_ABST
Abstract
本申请提供了一种慢性心衰预测模型构建方法、装置、设备及存储介质，其中，该方法包括：获取生物医学文献样本数据、多病种临床样本数据以及目标病种临床样本数据；根据生物医学文献样本数据，对通用大语言模型进行训练，生成生物医学大语言模型；根据多病种临床样本数据，对生物医学大语言模型进行指令微调训练，生成临床生物医学大语言模型；根据目标病种临床样本数据，对临床生物医学大语言模型进行低秩适应微调训练，并将训练结束时的临床生物医学大语言模型作为慢性心衰预测模型。本申请能够在有限的真实临床数据样本情况下，使得模型预测准确率更高，稳定性更强，鲁棒性更强。
Need to check novelty before this filing date? Find Prior Art