基于超声图像的乳腺钙化点自动检测技术

Research on automatic detection of breast calcification point based on ultrasound image

当前乳腺钙化点检测方法多基于X光片,难以应用于超声图像,本研究提出基于超声图像的乳腺钙化点自动检测技术,首先将乳腺超声图像中的肿瘤区域通过勾画模板提取出来,基于简单线性迭代聚类算法进行超像素分割;然后提取表征各超像素的特征量来计算显著性图,基于钙化区域显著性进行粗钙化点分割;最终对分割后的粗钙化点进行形态学检测,达到对超声图像中的细钙化点自动检测。该方法取得了较好的分割效果,具有较强的鲁棒性,为形成具有普适性的肿瘤自动诊断方案奠定了研究基础。

It is important for early diagnosis of breast cancer to detect breast calcification points in ultrasound images.At present,the detection methods of calcification point are mostly based on mammography,which are difficult to apply in ultrasound images.We proposed an automatic detection technology of breast calcification point based on ultrasound image.First,the region of breast ultrasound image was extracted by sketching template,and then superpixel segmentation was carried out based on simple linear iterative clustering algorithm.Second,the saliency map was obtained by calculating the feature quantities representing each super-pixel,and the coarse calcified points were segmented based on the saliency of calcified regions.Finally,morphological detection of the rough calcification points after segmentation was carried out to achieve the automatic detection of fine calcification points in ultrasound images.This method has achieved good segmentation effect and strong robustness,which lays a foundation for the formation of a universal automatic cancer diagnosis scheme.