摘要
乳腺肿块分割的任务是在感兴趣区域内区分出肿块和正常组织。目前常用的病变分割算法主要包括区域生长等传统算法和全卷积网络等深度学习方法,这些算法都忽略肿块分割的一个特异性。和自然场景中明显的梯度边缘相比,乳腺肿块和背景的边缘是渐变的。标注的边缘不足够准确,训练过程中边缘周围错误标注的像素会影响网络的训练。因此,本文提出了基于腐蚀损失的肿块分割算法。利用形态学腐蚀操作生成计算损失的掩码,使错误可能性高的边缘周围部分不参与损失的计算,将其结合到全卷积加条件随机场的网络中。边缘模糊的特性存在于多种医学图像中,本文的方法具有普适性,可应用于其他医学病变的分割算法。
The breast mass segmentation task is to distinguish masses from normal tissues in the region of interest.At present,commonly used lesion segmentation algorithms mainly include traditional algorithms such as region growth and deep learning methods such as full convolutional networks.These algorithms all ignore the specificity of mass segmentation.Compared with the obvious edges in natural scenes,the edges of the breast mass are gradually changing.The annotated edges are not accurate enough.Pixels mislabeled around the edges will affect the training of the network during the training process.Therefore,this paper proposes a segmentation algorithm based on erosion loss.The morphological erosion operation is used to generate a mask for calculating the loss,so that the part around the edge with a high possibility of error does not participate in the loss calculation,and it is combined into a network with full convolution and conditional random field.The characteristics of blurred edges exist in a variety of medical images.The method in this paper is universal and can be applied to segmentation algorithms for other medical lesions.
作者
李响
卜巍
邬向前
LI Xiang;BU Wei;WU Xiangqian(School of Computer Science and Technology,Harbin Institute of Technology,Harbin 150001,China;School of Media Technology and Art,Harbin Institute of Technology,Harbin 150001,China)
出处
《智能计算机与应用》
2020年第6期48-50,共3页
Intelligent Computer and Applications
基金
国家自然科学基金(61672194)
国家重点研究与发展计划(2018YFC0832304)
中国黑龙江省杰出青年科学基金(JC2018021)
国家机器人与系统国家重点实验室项目(SKLRS-2019-KF-14)
中兴通讯产学研合作论坛合作项目。
关键词
肿块分割
腐蚀损失
边缘模糊
Mass segmentation
Erosion loss
Blurred edges
