基于决策树的海量医学图像数据挖掘方法研究  

Research on data mining method of mass medical image based on decision tree

作　　者：任仪[1] REN Yi(Henan Province People's Hospital,Zhengzhou 450000,China)

机构地区：[1]河南省人民医院,河南郑州450000

出　　处：《电子设计工程》2019年第6期33-36,41,共5页Electronic Design Engineering

基　　金：国家自然科学基金项目(51175084)

摘　　要：传统数据挖掘方法不能清晰描述医学图像的特征信息和高分辨属性。为解决上述问题,提出一种基于决策树的海量医学图像数据挖掘方法。分析图像的简化属性分裂特点,并通过引入矫正函数的方式定义决策树分类堆,完成海量医学图像挖掘决策树分类规则的确定。在此基础上,利用医学图像目标识别和灰度直方特征提取结果,确定局部数据的挖掘引子,完成基于决策树的海量医学图像数据挖掘方法研究。模拟方法运行环境设计对比实验结果表明,与传统数据挖掘方法相比,应用基于决策树的海量医学图像数据挖掘方法后,医学图像特征信息、高分辨属性的描述清晰程度均得到20%左右的提升。Traditional data mining methods cannot clearly describe the feature information and high resolution attributes of medical images.In order to solve the above problems,a mass medical image data mining method based on decision tree is proposed.This paper analyzes the characteristics of simplified attribute segmentation of images,and defines the decision tree classification heap by introducing correction function,so as to determine the decision tree classification rules of massive medical image mining.On this basis,the use of medical image target recognition and gray feature extraction results,to determine the local data mining teases,complete medical image data mining method based on decision tree massive research.Simulation running environment design contrast experiment results show that compared with traditional data mining methods,application of huge amounts of medical image data mining method based on decision tree,the description of the characteristics of medical image information and high resolution properties and clarity,averaging about 20%.

关 键 词：决策树 数据挖掘 简化分裂 矫正函数 目标识别 灰度直方特征 

分 类 号：TN014[电子电信—物理电子学]