摘要
目的为B超诊断乳腺肿瘤建立计算机辅助诊断手段,以降低活检数以及提高诊断的准确性和客观性。方法通过提取良性和恶性肿瘤B超图像的形态特征和灰度特征,包括傅立叶描述子,粗糙度和前后场回声比,组成特征矢量,再用k-均值聚类算法对特征矢量进行分类处理。结果k-均值聚类算法对良性肿瘤的识别率为89.85%,对恶性肿瘤的识别正确率达78.26%。结论本文中建立的方法能较肉眼更精确地反映良性和恶性肿瘤B超图像的特征,如果再结合医生的临床经验能大大提高乳腺肿瘤的诊断准确性。
Objective To provide a computer-aided method for the diagnosis of breast tumor by B-mode ultrasonic imaging. Methods The shape, margin, and intensity features including Fourier descriptor, roughness, and ratio of mean intensity were calculated from B-mode ultrasonic benign tumor and malignant tumor images. Feature vectors which indicated two clas ses of images were created with the three features. Then we used k-means clustering algorithm to classify vectors. Results The accuracy rates of k-means clustering algorithm were 89.85% for benign and 78.26% for malignance. Conclusion This technology could indicate the characteristics of B-mode images of benign tumor and malignant tumor more accurately than eyes did. tt could greatly improve the diagnostic accuracy of breast tumor.
出处
《中国医学影像技术》
CSCD
北大核心
2005年第11期1758-1760,共3页
Chinese Journal of Medical Imaging Technology
基金
四川省青年科技基金(05ZQ026-019)
四川省应用基础研究项目(03JY029-072-2)资助
关键词
乳腺肿瘤
傅立叶描述子
灰度特征
图像识别
Breast neoplasms: Fourier descriptor
Intensity feature
Image recognition