|
|||||
|
|
| 一种用于寻找典型病理图像的聚类分析方法 | |
| 引用本文： | 关鹏,全宇,何苗,周宝森. 一种用于寻找典型病理图像的聚类分析方法[J]. 中国医科大学学报, 2006, 35(5): 500-501 |
| 作者姓名： | 关鹏 全宇 何苗 周宝森 |
| 作者单位： | 1. 中国医科大学公共卫生学院流行病学教研室,辽宁,沈阳,110001 2. 中国医科大学公共卫生学院附属第二医院计算机室,辽宁,沈阳,110001 3. 中国医科大学公共卫生学院附属第一医院计算机室,辽宁,沈阳,110001 |
| 摘 要： | 目的：探讨利用K—medoids聚类分析方法来寻找典型病理图像。方法：对正常、低度和高度鳞状上皮内病变宫颈细胞的51个特征参数进行标准化处理，以消除原始变量数量及量纲影响。运用统计软件R对所有特征参数进行K—medoids聚类分析。结果：对于正常、低度和高度鳞状上皮内病变宫颈细胞的分类正确率分别为91．7％、98．0％和92，0％。作为类中心点的3个病理图片可作为典型病理图片指导临床及教学研究。结论：该方法简单、方便，客观性强，可应用到其他医学领域定量研究。 |
| 关 键 词： | K.medoids聚类 典型图像 |
| 文章编号： | 0258-4646(2006)05-0500-02 |
| 收稿时间： | 2006-02-21 |
| 修稿时间： | 2006-02-21 |
A clustering method for searching typical pathological images |
|
| GUAN Peng,QUAN Yu,HE Miao,ZHOU Bao-sen. A clustering method for searching typical pathological images[J]. Journal of China Medical University, 2006, 35(5): 500-501 | |
| Authors: | GUAN Peng QUAN Yu HE Miao ZHOU Bao-sen |
| Affiliation: | 1. Department of Epidemiology, College of Pubhc Health, China Medical University, Shenyang 110001, China; 2. Computer Center, The Second Affdiated Hospital; 3. Computer Center, The First Affiliated Hospital |
| Abstract: | Objective: To investigate the potentiality of K-medoids cluster analysis in searching typical pathological images.Methods: All 51 characters of the samples from normal cervical cells,cervical cells in low-grade squamous intraepithelial lesion(LSIL),and cervical cells in high-grade squamous intraepithelial lesion(HSIL) were standardized for eliminating the influence of the size and dimension of original variables.Statistical software R was used to conduct K-medoids cluster analysis.Results: The diagnostic accuracy of normal,LSIL,and HSIL cervical cells were 91.7%,98.0%,92.0%,respectively.The medoids identified from each cluster can be used as typical images for teaching and research.Conclusion: K-medoids cluster analysis is an easy,convenient,objective method for searching typical pathological images and can be used in medical quantitative studies. |
| Keywords: | K-medoids clustering typical image |
| 本文献已被 CNKI 维普 万方数据 等数据库收录！ | |