基于深度学习的肝脏肿瘤图像分割  

Liver Tumor Image Segmentation Based on Depth Learning

作　　者：黄佳佳 赵曙光[1] 张笑青 杨峰 许方成 Huang Jiajia;Zhao Shuguang;Zhang Xiaoqing;Yang Feng;Xu Fangcheng(School of Information Science and Technology,Donghua University,Shanghai 201620)

机构地区：[1]东华大学信息科学与技术学院,上海201620

出　　处：《数字技术与应用》2017年第11期81-82,共2页Digital Technology & Application

摘　　要：传统肝脏CT图像肿瘤分割算法常需引进先验信息,分割效率低,无法满足临床实时性要求。针对这些问题,提出基于卷积-反卷积神经网络的肝脏肿瘤图像分割算法,自动提取肝脏肿瘤的特征。实验表明,DSC指标约85.32%,且分割每幅图像只需用时3至5秒,实时性好。The traditional liver CT image segmentation algorithm often needs to introduce prior information,and the segmentation efficiency is low,which can not meet the real-time clinical requirements.To solve these problems,an algorithm of hepatic tumor segmentation based on convolutiondeconvolution neural network is proposed to automatically extract features of liver tumors.Experiments show that the DSC index is about85.32%,and the segmentation of each image takes only3to5seconds,good real-time.

关 键 词：肝脏CT图像 肿瘤分割 反卷积网络 自动提取特征 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]