基于元学习、注意力机制、LSTM等认知智能方法的慢性病个性化用、调药预测技术和方法研究

针对高血压治疗周期长、治疗方案多样的特点，本项目面向高血压患者个性化治疗领域，采用大数据挖掘、深度学习等技术和方法，以为高血压患者提供个性化用药方案为目标，基于申请人所在课题组积累的海量高血压患者临床数据，深入研究个性化用药推荐和用药调整预测问题。首先，通过收集、组织、整理现有的医学知识库，构建以用药方案为中心的，涵盖多源异构数据关联关系的医疗知识图谱；其次，提出基于LSTM的个性化用药推荐模型，利用多任务学习推荐药物组合方案；最后，基于关系增强型层级多任务学习的疾病预测，构建基于双序列交互学习的药物调整模型。目标是在我国实施健康中国战略的背景下，突破相关关键技术，实现智能决策支持在典型慢性病的个性化诊疗中的应用。

Based on the massive medical data of hypertensive patients accumulated by the applicant's research group, this project aims at providing personalized medicine recommendation and medicine adjustment prediction for hypertensive patients by using big data mining and deep learning methods. Firstly, by collecting, analyzing and organizing the existing medical knowledge base, we construct a medical knowledge graph centered on medication strategies and covering the association relationships of multi-source heterogeneous data; secondly, we propose a personalized medication recommendation model based on LSTM and use multi-task learning to recommend medication combination strategies; finally, we build a dual-sequence interaction learning-based medication adjustment model. The goal is to break through the relevant key technologies and realize the application of intelligent decision support in personalized diagnosis and treatment of typical chronic diseases in the context of implementing the health China strategy.