| 柯丽,张雪,康雁.基于分形特征序列的乳腺X线图像分类方法[J].中国医学影像技术,2012,28(3):582~586 |
| 基于分形特征序列的乳腺X线图像分类方法 |
| Classification of mammography based on fractal features sequence |
| 投稿时间：2011-07-18 修订日期：2011-10-28 |
| DOI： |
| 中文关键词: 乳腺病变 计算机辅助诊断 分形维数 多级分形特征 特征选择 支持向量机 |
| 英文关键词:Breast diseases Computer-aided diagnosis Fractal dimension Multi-level fractal features Feature selection Support vector machine |
| 基金项目: |
|
| 摘要点击次数: 2849 |
| 全文下载次数: 882 |
| 中文摘要: |
| 目的 表征乳腺图像中肿块部分纹理特征,通过纹理分析实现乳腺图像中肿块部分与正常腺体部分的分类。方法 应用分形特征值表征乳腺图像纹理特征,利用多级分形特征提取法将乳腺图像分解成一系列细节图像,提取出多个分形特征值;利用分类精度、ROC曲线及曲线下面积(AUC)进行特征选择构建分形特征序列,最后应用支持向量机(SVM)方法进行分类。结果 对60幅图像的可疑病变区域进行分形特征序列提取分析,SVM交叉验证分类精度达84.50%。结论 基于分形维数的乳腺图像分类方法不仅能对肿块与正常腺体进行图像分类,还可有效表征乳腺图像的纹理信息,有助于提高乳腺肿块诊断的准确率。 |
| 英文摘要: |
| Objective To describe the texture features of mass and implement the classification for breast mass and normal glands by texture analysis. Methods The texture features were described using fractal dimension. The detailed breast images were obtained by multi-level fractal features extraction methods of breast mass, and many features were extracted. The detectable rate, ROC curve and area under curve (AUC) were utilized to establish the fractal feature vector, and then breast images were classified using support vector machine (SVM) method. Results Sixty suspicious areas were extracted and classified, and the SVM cross-validation accuracy was 84.50%. Conclusion The breast image classification methods based on the fractal dimensions can classify the breast mass and normal gland, describe the texture features of mammograms efficiently, and improve the accuracy rate for mass detection. |
| 查看全文 查看/发表评论 下载PDF阅读器 |
|
|
|
