基于图像融合的乳腺肿瘤感兴趣区域边缘识别  

Edge recognition of breast tumor region of interest based on image fusion

作　　者：曾果 刘彦荣[1] 王力[1] 许方彧 蒋烈夫[2] 靳玉川[3] ZENG Guo;LIU Yanrong;WANG Li;XU Fangyu;JIANG Liefu;JIN Yuchuan(Medical Imaging Center,Nanyang Second General Hospital,Nanyang 473000,China;Nanyang Medical College,Nanyang 473000;Hebei Medical University,Shijiazhuang 050017,China)

机构地区：[1]南阳市第二人民医院医学影像中心,南阳473000 [2]南阳医学高等专科学校,南阳473000 [3]河北医科大学,石家庄050017

出　　处：《生物医学工程研究》2020年第4期337-341,共5页Journal Of Biomedical Engineering Research

基　　金：河南省中医管理局项目(2018ZY1007)。

摘　　要：目前,乳腺肿瘤感兴趣区域(region of interest,ROI)边缘的识别方法中,单一的病变检查图像无法全面反映出肿瘤情况,导致识别的准确率不足。针对该问题,本研究提出一种基于图像融合的乳腺肿瘤感兴趣区域边缘识别方法。首先运用加权平均图像融合技术融合不同设备采集的病理图像,然后采用Normalized Cut法提取图像的肿瘤边缘。利用核极限学习算法,建立肿瘤感兴趣区域模型后,输入肿瘤边缘得出肿瘤特征因素,最后使用ROI技术实现乳腺肿瘤感兴趣区域边缘识别。对比验证表明,本研究方法的识别准确率更高,具有可行性。In the region of interest(ROI)edge recognition method for breast tumors,the single inspection lesion image can not fully reflect the tumor,which causes insufficient recognition accuracy.In view of this problem,we proposed an edge recognition method of breast tumor region of interest based on image fusion.The pathological images collected by different devices were fused by using weighted averange image fusion technology,after the ROI model of tumor was established by using kernel limit learning algorithm,the tumor edge was inputted to get the tumor feature factors.Finally,ROI technique was used to realize the edge recognition of the breast tumor region of interest.Comparative verification shows that this method has higher recognition accuracy,and it is feasible.

关 键 词：乳腺肿瘤 边缘识别 准确率 图像融合 加权平均 

分 类 号：R318[医药卫生—生物医学工程] R814.3[医药卫生—基础医学]