«上一篇/Previous Article|本期目录/Table of Contents|下一篇/Next Article»

《计算机技术与发展》[ISSN:1006-6977/CN:61-1281/TN]

卷:

34

期数:

2024年10期

页码:

156-163

栏目:

人工智能

出版日期:

2024-10-10

文章信息/Info

Title:

A Clinical Pathway Mining Approach Based on 3D Sub-trajectory Clustering Algorithm

文章编号:

1673-629X(2024)10-0156-08

作者:

刘玉江1; 罗双红2; 3; 郑庆霄1*

1. 成都信息工程大学 区块链产业学院,四川 成都 610225;2. 四川大学华西第二医院,四川 成都 610066;3. 四川省儿科质量控制中心,四川 成都 610066

Author(s):

LIU Yu-jiang1; LUO Shuang-hong2; 3; ZHENG Qing-xiao1*

1. School of Blockchain Industry,Chengdu University of Information Technology,Chengdu 610225,China;2. West China Second University Hospital,Sichuan University,Chengdu 610066,China;3. Pediatric Quality Control Center of Sichuan Province,Chengdu 610066,China

关键词:

数据挖掘; 临床路径; TRACLUS算法; 轨迹聚类; KD树; 时间加权

Keywords:

data mining; clinical pathways; TRACLUS algorithm; trajectory clustering; K-dimension tree; time weighting

分类号:

TP301

DOI:

10.20165/j.cnki.ISSN1673-629X.2024.0188

摘要:

针对临床路径的制定和实施存在受医院自身条件及疾病复杂程度等因素影响的问题,提出一种基于三维子轨迹聚类算法的临床路径挖掘方法。 根据临床诊疗过程数据具有规律性和时序性的特征,该方法首先将大量临床诊疗过程数据转换为近似航空轨迹的三维时序轨迹,并使用时间加权方法对轨迹分布进行调整。 其次基于传统 TRACLUS 算法,使用 KD 树进行邻近搜索加速优化,引入希尔伯特空间中的相似性度量方法使其适应于高维轨迹聚类。 最后通过对大量轨迹进行聚类分析,从中提取出典型的临床诊疗过程,进而得到实际实施的临床路径。 通过使用该方法对某三甲医院新生儿低血糖临床数据进行了一系列实验,结果证实该方法能够提炼出本地各种情况下实际实施的临床路径,可辅助医生制定更个性化的治疗方案,并且该实验结果为标准新生儿低血糖临床路径的改进和实施提供了方向和有力依据。

Abstract:

Aiming at the problem that the formulation and implementation of clinical pathways are affected by the hospital’s own conditions and the complexity of diseases,we propose a clinical pathway mining method based on the three-dimensional sub-trajectory clustering algorithm. According to the characteristics of clinical diagnosis and treatment process data with regularity and temporal sequence,the proposed method first converts a large amount of clinical diagnosis and treatment process data into three-dimensional time-sequential trajectories that approximate the flight trajectories,and adjusts the distribution of trajectories using time-weighted methods.Second,based on the traditional TRACLUS algorithm,K-dimension tree is used to accelerate the optimization of neighborhood search,and the similarity measure in Hilbert space is introduced to make it adaptable to high - dimensional trajectory clustering. Finally, by clustering and analyzing a large number of trajectories,a typical clinical diagnosis and treatment process is extracted from them,and then a clinical pathway is obtained for actual implementation. A series of experiments were conducted on the clinical data of neonatal hypoglycemia in a tertiary hospital, and the results confirmed that the proposed method can extract the clinical pathways actually implemented in various local situations,which can help doctors formulate a more personalized treatment plan,and the results of the experi-ments provide a direction and a strong basis for the improvement and implementation of the standard clinical pathways for neonatal hypo-glycemia.

常用功能

更新日期/Last Update: 2024-10-10