农业图书情报学报2025，Vol.37Issue(4)：24-38,15.DOI:10.13998/j.cnki.issn1002-1248.25-0079
多模态医养大数据动态聚合与智慧服务模式
Big Data Dynamic Aggregation and Intelligent Service Model for Multimodal Healthcare and Eldercare
摘要
Abstract
[Purpose/Significance]Against the backdrop of an accelerating population aging trend,the integration of big data and intelligent services in multimodal healthcare and eldercare has become pivotal for enhancing the quality of medical and eldercare services.However,existing knowledge service systems for big data in healthcare and eldercare face challenges such as difficulty of integrating multi-source heterogeneous data,the absence of cross-organizational sharing mechanisms,and passive service models.[Method/Process]First,a cross-domain aggregation method is proposed for multi-source heterogeneous medicare big data,including:1)A method for constructing a clinical,key-feature-based medical case knowledge database.It extracts and categorizes critical features from electronic medical records using natural language processing(NLP).2)A natural language processing-based cross-domain disease risk factor mining framework.It identifies risk factors from social media via topic-enhanced word embeddings and clustering techniques.3)An adaptive pointer-constrained generation method for medical text-to-table tasks.It leverages the BART architecture to transform unstructured medical text into structured tables.Next,a knowledge discovery method based on multimodal medicare big data is developed,including:1)A medical decision support approach integrating case-based reasoning(CBR)and explainable machine learning.It aims to enhance diagnostic interpretability through ensemble learning and case similarity analysis.2)A large-scale medical model-driven knowledge system.It utilizes multimodal data pretraining and domain adaptation to support the entire diagnosis-treatment process.3)A personalized recommendation method based on temporal warning signals,generating precise intervention plans via collaborative filtering and dynamic updates.Finally,a smart service model for full-cycle evolving needs is constructed,including:1)A health information supply-demand consistency matching framework combining deep learning and clustering techniques;2)A multi-level,cross-scenario health demand and behavior dynamic modeling approach.[Results/Conclusions]The proposed methodological framework significantly improves the efficiency with which medicare big data is integrated and the capabilities of its knowledge services.Key outcomes include:1)Enabling disease risk prediction and personalized interventions through deep integration of cross-organizational,cross-scenario medicare data via multimodal aggregation and semantic alignment.2)The CBR-ECC model and WiNGPT large medical models enhance the interpretability and full-process coverage of medical decision-making.These models improve the accuracy of diagnoses made by primary care physicians by over 30%.3)The temporal warning-based recommendation method increases the dynamic update efficiency of health interventions by 40%and user satisfaction by 25%;4)Dynamic health demand modeling reveals core pain points for chronic disease patients,providing a basis for precision service strategies.This research provides the theoretical and technical support for developing a proactive health service system that is both data-driven and human-machine collaborative.This system will,advance the implementation of the Healthy China strategy and innovation in aging population governance.关键词
多模态医养大数据/深度聚合/智慧服务/人机协同/健康知识图谱Key words
big data in multimodal healthcare and eldercare/deep integration/intelligent services/human-machine collaboration/health knowledge graph分类
社会科学引用本文复制引用
杨雪洁,刘佳,吴青筱,王雨菲,顾东晓..多模态医养大数据动态聚合与智慧服务模式[J].农业图书情报学报,2025,37(4):24-38,15.基金项目
国家自然科学基金面上项目"医防融合的疾病多渠道协同防治与智能管理模式研究"(72271082) （72271082）
"基于多模态医养大数据深度聚合的动态主动服务模式研究"(72071063) （72071063）
安徽省自然科学基金杰青项目"医联网环境下的医疗健康知识挖掘与智能决策方法研究"(2408085J041) （2408085J041）
合肥工业大学学术新人提升计划A项目"数字变革对医疗服务提供的影响研究"(JZ2025HGTA0169) （JZ2025HGTA0169）
