基于LBP和极限学习机的脑部MR图像分类

发布时间：2018-05-18 10:48

本文选题：MR图像 + 局部二值模式　；参考：《山东大学学报(工学版)》2017年02期

【摘要】：为解决磁共振(magnetic resonance,MR)脑部图像来源不一以及病变位置和形态不固定造成MR脑部图像分类精度不高的问题,提出基于局部二值模式(local binary pattern,LBP)的纹理特征提取,并用极限学习机(extreme learning machine,ELM)对M R图像分类。计算图像感兴趣区域(region of interest,ROI)的掩码,将图像分成扇形的子区域,统计掩码坐标下各块子区域的LBP直方图,连接所有LBP直方图作为特征向量通过ELM进行分类。相比以前的方法,该方法能够计算颅脑内局部纹理特征,能分类来源不一以及多种病变的图像。对脑部M R图像分类进行试验,对所有样本分类正确率超过92%,正类样本正确率超过93%,负类样本正确率超过91%。试验结果表明,该方法能够对较为复杂的MR图像进行正确分类。
[Abstract]:In order to solve the problem of low classification accuracy caused by the different sources of magnetic resonance MRI brain images and the location and shape of the lesions, a method of texture feature extraction based on local binary mode (local binary pattern LBP) is proposed to solve the problem that the classification accuracy of Mr images is not high due to the location and shape of the lesions. M R images were classified with extreme learning machine ELM. The mask of region of interest in the image is calculated, and the image is divided into sector subregions. The LBP histograms of each sub-region in the statistical mask coordinates are connected to all LBP histograms as feature vectors to be classified by ELM. Compared with the previous methods, this method can calculate the local texture features of the brain, and can classify images with different sources and various lesions. The classification accuracy of all samples is more than 92%, the correct rate of positive samples is more than 933%, and the accuracy rate of negative samples is more than 91%. The experimental results show that this method can correctly classify the more complicated Mr images.
【作者单位】：桂林电子科技大学电子工程与自动化学院;
【基金】：国家自然科学基金资助项目(61105004) 广西高校图像图形智能处理重点实验室基金资助项目(LD16096X) 桂林电子科技大学创新基金资助项目(GDYCSZ201428)
【分类号】：R445.2;TP391.41

【相似文献】