基于核密度估计聚类和关联规则的医学图像分类

A Kernel Density Estimation and Association Rule-based Medial Image Classification Algorithm

基于数据挖掘的医学图像分类方法研究是多媒体数据挖掘的一个重要组成部分。在分析和总结了现有各种特征提取方法的基础上,提出了基于核密度估计聚类和关联规则的医学图像分类算法和关联规则的医学图像分类器框架。该算法先用核密度估计的聚类算法实现医学图像的聚类,在聚类的结果上提取局部特征,在局部特征上用关联规则实现医学图像的分类。实验结果表明可以较好的提高医学图像分类的准确率。

Data mining based medical image classification methods is an important component of multimedia data mining. Found on analysis and summary of feature extraction methods, this paper presents a new medical image classification algorithm based on kernel density estimation and association role. Firstly, this method uses the kernel density estimation - clustering algorithm to cluster medical image. Then, local features are extracted on the clustered results and association rules are used to classify medical images based on those local features. Experiment results indicate that this method can obviously improve the accuracy of medical image classification.