摘要
为了对冠状动脉图像的精确分割,提出了一种基于多源融合FCN的冠脉图像分割方法研究.首先对原始CTA图像进行双边滤波和LOG算子处理,并将处理后的图像与原始图像共同构成多源图像作为输入,将传统全卷积网络拓展成具有多源融合特色的分割模型.通过一系列的实验表明了全卷积神经网络在冠状动脉图像分割上的有效性.
In order to accurately segment the coronary artery image,a coronary image segmentation method was proposed based on multi-source fusion FCN.In this method,the original CTA image was firstly processed by bilateral filter and LOG operator;then the processed image and the original image were combined to form a multi-source image as the input;finally,the traditional full convolution network was expanded into a segmentation model with multi-source fusion characteristics.The effectiveness of the full convolutional neural network in coronary artery image segmentation was indicated through a series of experiments.
作者
段军
常一凡
DUAN Jun;CHANG Yifan(Mining Research Institute,Inner Mongolia University of Science and Technology,Baotou 014000,China;Information Engineering School,Inner Mongolia University of Science and Technology,Baotou 014000,China)
出处
《内蒙古科技大学学报》
CAS
2019年第3期277-282,共6页
Journal of Inner Mongolia University of Science and Technology
基金
国家自然科学基金资助项目(61663036)
关键词
冠状动脉分割
全卷积神经网络
多源融合
多尺度特征
深度可分卷积
coronary artery segmentation
full convolutional neural network
multi-source fusion
multiscale feature
depthwise separable convolutions
