基于3D卷积网络的结肠癌CT图像分类研究  

Study of Colorectal Cancer CT Image Classification Based on 3D Convolutional Network

作　　者：吕刚[1] 应明亮[2] LYU Gang;YING Mingliang(Graphic Information Center,Jinhua Radio and Television University,Jinhua 321000,Zhejiang,P.R.China;Department of Radiology,Jinhua Municipal Central Hospital,Jinhua 321000,Zhejiang,P.R.China)

机构地区：[1]金华广播电视大学图文信息中心,浙江金华321000 [2]金华市中心医院放射科,浙江金华321000

出　　处：《影像科学与光化学》2022年第5期1024-1028,共5页Imaging Science and Photochemistry

基　　金：浙江省医药卫生科技项目(2022KY1330);金华市科技局重点项目(2020-3-037)。

摘　　要：结直肠肿瘤是全世界排名前三的肿瘤之一,目前在CT图像中筛查出结肠肿瘤仍需要放射科医生手工完成,这是一项费时费力的重复性劳动,文中研究了用3D卷积网络在腹部CT图像上提取3D特征,判断CT图像中是否有结肠癌病灶的方法。文中采集了一个有348例样本的结肠癌CT图像数据集,设计了3种不同结构的3D卷积网络,对采集的腹部CT图像进行了正常/非正常二分类实验,文中性能最好的模型C是一个带有3个改进的残差模块的3D卷积网络,正常/非正常二分类实验的平均准确率为96.2%,AUC为0.989,分别比基准模型提高了2.2%、2.9%。实验结果表明,3D卷积网络在结肠癌CT的正常/非正常二分类任务上准确率高,泛化能力好,且只需要CT级弱标定数据集,具有潜在的临床应用价值。Colorectal cancer(CRC) is among the top three tumors in the world. At present, screening CRC in CT images still needs to be done manually by radiologists, which is a time-consuming and laborious repetitive work. This paper investigated a method which extracted 3 D features on abdominal CT images based on 3 D convolutional network, then estimated whether there were CRC lesions through CT images. This paper collected a colorectal cancer CT image dataset with 348 samples and designed three 3 D convolution networks with different structures, and performed a normal/abnormal binary classification experiments on the acquired abdominal CT images. The best performance model in this paper was a 3 D convolution networks with three improved residual modules. The average accuracy of normal/abnormal binary classification experiment was 96.2%, and the AUC was 0.989, which was 2.2% and 2.9% higher than the baseline model. The experiment results showed that the 3 D convolutional networks with have outstanding performance in the CRC normal/abnormal classification task, it also has good generalization ability and only under CT-level weak-labeled data, it is helpful to clinical application.

关 键 词：3D卷积 残差学习 CT图像分类 深度学习 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程] R735.35[医药卫生—肿瘤]