摘要
目的:通过深入分析乳腺病理图像,为精确区分良恶性乳腺肿瘤,提出便于计算又能给出较高的分类精度的纹理特征参数。方法:基于灰度共生矩阵算法,提取乳腺癌病理图像的纹理特征进行分析。结果:确定4个具有很好特征效果且便于计算的纹理特征参数,熵和对比度的均值特征对区分良恶性肿瘤有很好的表现。结论:在乳腺病理图像中提取熵和对比度的均值为主要特征,可有效区分乳腺肿瘤良性与恶性。
Objective: To accurately distinguish benign and malignant breast tumors and indicate the texture parameters that could conveniently calculate and could provide higher classification accuracy through in-depth analysis for breast pathology images. Methods: Based on the gray scale co-occurrence matrix algorithm, the texture features of pathological imagesof breast tumorwere extracted and analyzed. Results:There were 4 parameters of texture feature with good feature effects and easy calculation were determined. And the features of mean value of entropy and contrast ratio have good performance in distinguishing benign and malignant tumors. Conclusion:Through extracts the mean values of entropy and contrast ratio as main features in the pathological images of breast, the benign and malignant breast tumors can be effectively distinguished.
作者
赵爽
李延军
马志庆
赵文华
ZHAO Shuang;LI Yan-jun;MA Zhi-qing(Polytechnic College,Shandong University of Traditional Chinese Medicine,Jinan 250355,China)
出处
《中国医学装备》
2018年第8期5-8,共4页
China Medical Equipment
基金
山东省研究生教育创新计划(SDYY16069)"基于<生物医学图像处理与分析>课程群移动学习资源的研究与设计"
关键词
灰度共生矩阵
乳腺癌病理图像
纹理
特征提取
熵和对比度
Gray co-occurrence matrix
Breast pathology image
Texture
Feature extraction
Entropy and contrast ratio