论文部分内容阅读
目的 通过纹理分析的方法对脑胶质瘤的磁共振图像(包括瘤区和瘤周)进行分析,并用支持向量机(SVM)评价纹理特征.方法 选取24例高级别脑胶质瘤的感兴趣区(ROI)进行纹理分析,计算灰度图像统计信息特征,采用灰度共生矩阵提取受试高级别脑胶质瘤的增强T～WI磁共振图像,用软件Image J提取肿瘤的ROI,针对显示肿瘤最大层面的三层图像者的对比度、相关、能量、逆差矩和熵等纹理特征,并比较各特征之间的显著性差异.利用支持SVM结构风险最小化理论的优势进行分类器设计.结果 利用灰度共生矩阵提取出的5个纹理参数中至少有3个参数在上述两种不同类别的数据集之间具有统计学意义(P＜0.05),对于所有不同类别的数据集,ROI区域灰度值的方差参数均具有统计学意义.脑胶质瘤肿瘤区域和肿瘤周边区域具有显著性差异特征的SVM测试准确率为(90.72±2.27)％.结论 纹理特征的分析可以提供更多量化信息特征,为精确界定高级别脑胶质瘤肿瘤区域和肿瘤周边区域提供了一种新方法.“,”Objective To use the texture analysis method to analyze magnetic resonance images(control tumor area and peri-tumoral region) of gliomas,the texture characteristics of gliomas were evaluated by a support vector machine (SVM).Methods 24 patients with high grade gliomas were retrospectively analyzed.Enhanced T1 magnetic resonance images of the high grade gliomas were collected.The region of interest (ROI) was selected on three slices per patient in Image J,using thresholding and manual outlining.The statistical texture features of the gray image to compare the significant differences between the characteristics were calculated.Support vector machine (SVM) structure risk minimization theory was used to design a classifier.Results The 5 texture parameters extracted from the gray level co-occurrence matrix have at least 3 parameters which have statistical significance between the two different classes of data sets(P ＜ 0.05).For all kinds of data sets,the variance parameters of gray value of ROI region have statistical significance.The accuracy rate of SVM test was 90.73 ± 2.27 in the remarkable difference of tumor area and peri-tumoral regions of gliomas.Conclusion The analysis of texture features can provide more quantitative information,which provides a new idea and method for the precise definition of control tumor area and peri-tumoral region of the high grade gliomas.