改进的双水平集心脏MRI图像左心室分割算法  

An Improved Double Level Set Algorithm for Left Ventricular Segmentation of Cardiac MRI Images

作　　者：赵昊宸 苑金辉 朱恩嵘 乔艳 胡晓飞[2] ZHAO Hao-chen;YUAN Jin-hui;ZHU En-rong;QIAO Yan;HU Xiao-fei(School of Communication and Information Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210003,China;School of Geography and Bioinformatics,Nanjing University of Posts and Telecommunications,Nanjing 210003,China)

机构地区：[1]南京邮电大学通信与信息工程学院,江苏南京210003 [2]南京邮电大学地理与生物信息学院,江苏南京210003

出　　处：《计算机技术与发展》2022年第6期162-166,共5页Computer Technology and Development

基　　金：国家自然科学基金(61771251);江苏省社会发展重点项目(BE2016773)。

摘　　要：针对基于深度学习的心脏MRI图像左心室分割网络仅使用简单的阈值法对输出的概率图进行二值化,使得分割的结果边缘模糊的问题,提出了一种改进的基于双水平集的后处理算法。在使用传统水平集方法对左心室内外膜进行分割时,存在水平集函数演化不稳定、边缘分割精度低和分割效率低等问题。提出新的距离规则化函数对水平集的能量函数进行改进,新的距离规则化函数可以更好地约束边缘正则性,能够很好地弥补深度网络输出结果当中的缺陷,并使用双水平集,即0水平集函数和k水平集函数分别向左心室内外膜演化。利用多伦多市儿童病医院影像科提供的数据集,用上述算法对基于深度网络的心脏MRI图像左心室分割结果进行后处理,实验结果表明,左心室内膜和外膜分割的Dice相似系数分别为0.9338和0.9587。对比其他分割模型,分割精度获得了明显提高。Aiming at the problem that the edge of the ventricular segmentation results are fuzzy only using a simple threshold method to binarize the output probability map of the left ventricular segmentation network of cardiac MRI images based on deep learning,an improved post-processing algorithm based on two-level sets is proposed.When the traditional level set method is used to segment the left ventricle endocardium and epicardium,there are problems such as unstable evolution of the level set function,low accuracy of edge segmentation,and low segmentation efficiency.A new distance regularization function is proposed to improve the energy function of the level set.The new distance regularization function can better constrain the edge regularity,and can make up for defects in the output of the deep learning network.Simultaneously,we use the two-level set algorithm,that is,the 0 level set function and the k level set,to evolve to the left ventricle endocardium and epicardium respectively.The data set provided by the imaging department of the Toronto Children’s Hospital is applied in this thesis.The aforementioned algorithm is used to post-process the left ventricular segmentation results of cardiac MRI images based on the deep network.The experimental results show that the Dice similarity coefficients of the left ventricular endocardium and epicardium segmentation are 0.9338 and 0.9587,respectively.Compared with other segmentation models,the segmentation accuracy of this thesis has been significantly improved.

关 键 词：水平集方法 距离规则化函数 后处理算法 左心室图像分割 深度学习 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]