基于数据湖的非结构化医疗数据高效检索方法OA
Efficient retrieval method for unstructured medical data based on data lake
随着智能医疗设备广泛应用,非结构化医疗数据呈现爆炸性增长,其高效检索与管理对医疗领域中数据挖掘、数据探索和人工智能应用具有重要意义,可辅助医生优化治疗决策.然而,现有数据检索方法注重数据实体语义表达能力,缺乏对检索算法优化,导致大规模数据检索效率低下.高维索引和分布式存储平台是提高检索效率的有效途径,但现存研究没有将二者有效结合,在检索效率和存储灵活性方面仍存在不足.针对上述挑战,本文提出一种基于数据湖的非结构化医疗数据高效检索方法 EMUD(efficient re-trieval method for unstructured medical data based on data lake),它使用新颖高维索引,结合分布式计算能力,显著提升非结构化医疗数据检索效率.同时,EMUD 通过集成数据湖灵活数据存储特性,实现非结构化医疗数据高效管理,为医疗前沿应用提供灵活数据支撑接口.实验结果表明,在使用较小存储空间条件下,EMUD 在检索精度、检索效率以及索引创建时间方面均优于其他高维索引.
With the widespread adoption of intelligent medical devices,unstructured medical data has grown explosively.Efficient retrieval and management of such data are essential for data mining,exploratory analysis,and artificial in-telligence applications in the medical domain,thereby assisting clinicians in optimizing treatment decisions.How-ever,existing data retrieval methods primarily focus on the semantic representation of data entities and lack the op-timization of retrieval algorithms,resulting in low efficiency of large-scale data retrieval.High-dimensional indexes and distributed storage platforms are effective approaches to improving retrieval efficiency,but prior research has not achieved their effective integration,resulting in shortcomings in both retrieval performance and storage flexibil-ity.To address the above challenges,this paper proposes EMUD(efficient retrieval method for unstructured medical data based on data lake),an efficient retrieval method for unstructured medical data based on data lake.It levera-ges a novel high-dimensional index combined with distributed computing capabilities to enhance the retrieval effi-ciency of unstructured medical data significantly.Meanwhile,EMUD integrates the flexible data storage character-istics of the data lake to achieve efficient management of unstructured medical data and provide a scalable data sup-port interface for cutting-edge medical applications.The experimental results demonstrate that EMUD outperforms other high-dimensional indexes regarding retrieval accuracy,retrieval efficiency,and index construction time under the condition of using less storage space.
王凯歌;王炜凡;胡振涛;金勇
河南大学人工智能学院 郑州 450046河南大学人工智能学院 郑州 450046河南大学人工智能学院 郑州 450046河南大学人工智能学院 郑州 450046
数据湖非结构化医疗数据高效检索高维索引
data lakeunstructured medical dataefficient retrievalhigh-dimensional index
《高技术通讯》 2026 (7)
724-730,7
河南省科技攻关计划(252102211041,262102211085)资助项目.
评论