[论文解读] Q-Map: clinical concept mining with phrase sense disambiguation.

Sheikh Shams Azam, Manoj Raju+2 more|arXiv (Cornell University)|Apr 30, 2018

Biomedical Text Mining and Ontologies参考文献 9被引用 2

一句话总结

Q-Map 是一种快速、可配置的临床概念抽取系统，通过利用经过整理的知识源，高效地从非结构化医疗文本中提取结构化信息。它在速度和可配置性方面优于 MetaMap，同时在临床笔记的概念检索中保持了高精度。

ABSTRACT

Over the past decade, there has been a steep rise in data driven analysis in major areas of medicine, such as, clinical decision support system, survival analysis, patient similarity analysis, image analytics etc. Also, there are various ongoing research efforts in the operational and financial fields using techniques such as demand forecasting, convex optimization. Most of the data used in these research applications are well-structured and available in numerical or categorical formats which can be used for experiments directly. On the opposite end, there exists a wide expanse of data that is intractable for direct analysis owing to its unstructured nature. These can be found in the form of discharge summaries, clinical notes, procedural notes which are in human written free text format and neither have any relational model nor any standard grammatical structure. An important step in utilization of these texts for such studies is to transform and process the data to retrieve structured information from the haystack of irrelevant data using information retrieval and data mining techniques. The unregulated format coupled with massive size of datasets makes the mining process a monumental task requiring robust algorithms supported by ample hardware resources and computing power. In this paper, we present Q-Map, which is a simple yet powerful system that can sift through these datasets to retrieve structured information aggressively and efficiently. It is backed by an effective mining algorithm based on curated knowledge sources, that is both fast and configurable. We also present its comparative performance with MetaMap, one of the most reputed tools for medical concepts retrieval.

研究动机与目标

• 解决从非结构化自由文本医疗记录（如出院小结和临床笔记）中提取结构化临床概念的挑战。
• 开发一种可扩展且高效的系统，能够在计算开销最小的情况下处理大规模非结构化临床文本。
• 通过提升速度和可配置性，改进现有工具（如 MetaMap）在临床概念识别中的表现，同时保持高精度。
• 通过将非结构化临床文本转化为结构化、可分析的信息，支持数据驱动的医学研究。

提出的方法

• Q-Map 采用基于知识源的方法，利用经过整理的医学术语库，将临床文本中的短语映射到标准化概念。
• 应用短语语义消歧技术，解决临床术语中的歧义，提高概念映射的精确度。
• 系统采用模块化架构，可根据特定应用场景或数据类型进行配置。
• 利用高效的索引和匹配算法，加速大规模数据集中的概念检索。
• 通过标准临床概念抽取基准测试，与 MetaMap 进行性能对比评估。

实验结果

研究问题

• RQ1 如何高效地将非结构化临床文本转化为可用于医学研究的结构化、可分析数据？
• RQ2 与 MetaMap 等成熟工具相比，Q-Map 在概念检索方面的表现如何？
• RQ3 短语语义消歧在多大程度上提升了临床概念抽取的准确性？
• RQ4 是否可以开发一种既快速又可配置的系统，同时不牺牲临床文本处理中的精度？

主要发现

• Q-Map 在处理大规模临床文本数据集方面相比 MetaMap 显示出显著更快的处理速度。
• 该系统在概念检索中保持了高精度，通过短语语义消歧有效解决了术语歧义问题。
• 其模块化和可配置的设计使其能够适应多种临床数据类型和研究需求。
• Q-Map 在支持临床数据分析流水线可扩展部署的同时，显著降低了计算开销。