摘要
医学图像危及器官自动分割是计算机辅助诊断中的重要组成部分,对辅助医生高质高效完成放射治疗有着极其重要的作用。胸腔CT图像对比度低,且各器官之间重叠交错、边界模糊,使得危及器官的精确分割具有较大的挑战性。提出一种多尺度特征感知的编码-解码网络模型(FA-Unet),实现胸腔CT图像危及器官的分割。针对胸腔中四类器官大小差异的问题,首先构建了输入感知模型,提取图像中各器官的多尺度特征。为了弥补编码与解码之间的语义鸿沟,在解码-编码中融入改进的inception模块。用空间金字塔卷积(ESP)与金字塔池化(PSP)模块代替传统的串行卷积运算,使得网络模型更为轻量化,在一定程度上缓解数据量不足带来的过拟合问题。采用一种联合Dice系数与交叉熵的损失函数训练分割网络,可解决胸腔CT图像中类别不平衡的问题。最后,在2019年ISBI发布的Seg THOR数据集上验证模型的有效性,该数据集共包括40例肺癌或霍奇金淋巴瘤患者的胸腔CT图像7 390张。实验结果表明,胸腔CT图像各器官分割的Dice系数分别为食道0.793 2、心脏0.935 9、气管0.854 9、主动脉0.889 0,Hausdorff距离分别为食道1.420 7、心脏0.212 4、气管0.627 3、主动脉0.887 0。结果表明,与同类型分割网络相比,模型可获得较好的分割性能,尤其在小目标器官的分割上取得竞争性优势。
Automatic segmentation of organs at risk( OARs) in medical images is an essential constituent of computer-aided diagnosis,and it plays a vital role in assisting doctors to completeradiotherapy with high quality and efficiency. There are some challenges in the accurate segmentation of OARs for thoracic CT images,including low intensity contrast,different organs with interlaced and overlap regions,and different structure without clear boundaries. In this paper,a multi-scale feature-aware encoding-decoding network( FA-Unet) was proposed to segment OARs in thoracic CT images.To address the problem of the size difference among four kinds of organs in the thoraciccavity,an input-aware module was designed to extract multi-scale features in four types of organs. In order to bridge the semantic gap between the encoding and decoding layers,the modified inception module was introduced to long-range skipconnections between the encoding part and the decoding part in our architecture. Furthermore,we replaced the traditional serial convolution operation with the efficient spatial pyramid( ESP) andpyramid spatial pooling( PSP) modules to make our network more lightweight and avoide over-fitting caused by insufficient data effectively. We formulated a novel loss function by combining Dice coefficient and cross entropy to train our network to resolve the class imbalance in thoracic CT images. Finally,we evaluated the effectiveness of our model on the SegTHOR data set released by ISBI in 2019,and the dataset includes 7 390 thoracic CTimages of 40 patients with lung cancer or Hodgkin’s lymphoma. Experimental results showed that Dice coefficient of each organ in thoracic CT image was 0.793 2 of esophagus,0. 935 9 of heart,0. 854 9 of trachea and 0. 889 0 of aorta. Hausdorff distances was 1.420 7 of esophagus,0.212 4 of heart,0.627 3 of trachea and 0.887 0 of aorta. Experimental results verified that our proposed model outperformed other state-of-the-arts on the segmentation results of OARs and achieved very competitive performance on small target organs.
作者
邓仕俊
汤红忠
曾黎
曾淑英
张东波
Deng Shijun;Tang Hongzhong;Zeng Li;Zeng Shuying;Zhang Dongbo(College of Automation and Electronic Information,Xiangtan University,Xiangtan 411104,Hunan,China;Key Laboratory of Intelligent Computing and Information Processing of Ministry of Education,Xiangtan University,Xiangtan 411105,Hunan,China)
出处
《中国生物医学工程学报》
CAS
CSCD
北大核心
2021年第6期701-711,共11页
Chinese Journal of Biomedical Engineering
基金
湖南省自然科学基金(2020JJ4588,2020JJ4090)
智能计算与信息处理教育部重点实验室(湘潭大学)开放课题(2020ICIP06)。
关键词
胸腔CT图像
危及器官分割
多尺度特征感知
FA-Unet
thoracic CT image
segmentation of organs at risk
multi-scale feature-aware
feature-aware encoding-decoding network(FA-Unet)