基于集成DE-NRS的肺部肿瘤影像组学计算机辅助诊断模型  

COMPUTER AIDED DIAGNOSIS MODEL OF LUNG TUMOR IMAGING HISTOLOGY BASED ON INTEGRATED DE-NRS

作　　者：任海玲 周涛 霍兵强[4] Ren Hailing;Zhou Tao;Huo Bingqiang(School of Public Health and Management,Ningxia Medical University,Yinchuan 750004,Ningxia,China;School of Science,Ningxia Medical University,Yinchuan 750004,Ningxia,China;Ningxia Key Laboratory of Intelligent Information and Big Data Processing,Yinchuan 750021,Ningxia,China;School of Computer Science and Engineering,North Minzu University,Yinchuan 750004,Ningxia,China)

机构地区：[1]宁夏医科大学公共卫生与管理学院,宁夏银川750004 [2]宁夏医科大学理学院,宁夏银川750004 [3]宁夏智能信息与大数据处理重点实验室,宁夏银川750021 [4]北方民族大学计算机科学与工程学院,宁夏银川750004

出　　处：《计算机应用与软件》2020年第5期156-163,204,共9页Computer Applications and Software

基　　金：国家自然科学基金项目(61561040);陕西省教育厅项目(2013JK1142)。

摘　　要：针对肺部肿瘤计算机辅助诊断存在假阳性高和经典RS容错性差、不能处理连续型数据等问题,结合DE提出基于集成NRS的肺部肿瘤计算机辅助诊断模型。对肺部肿瘤CT、PET、PET/CT医学影像图像中的病灶区域进行截取与分割,并进行特征提取得到特征库;基于DE与NRS构建属性约简模型,得到特征子集;基于SVM分类器,搭建肺部CT、PET、PET/CT个体分类器;采取相对多数投票准则构造集成学习模型。进行有效性和可行性实验:邻域大小delta确定实验,变异系数F、交叉系数CR、权重值w参数实验,集成实验,并进行对比。结果表明:该模型在整体性能上较好,识别精度达到99.72%,具有较好的鲁棒性和可扩展性。Aiming at the problem of high false positive rate,poor fault tolerance of classical RS and inability to process continuous data in computer-aided diagnosis of lung tumors,we propose a computer-aided diagnosis model of lung tumors based on integrated NRS in combination with DE.The lesion regions of lung CT,PET and PET/CT medical images were intercepted and segmented,and feature extraction was performed to obtain a feature database;the attribute reduction model was built based on DE and NRS to get the feature subset;individual classifiers of lung CT,PET and PET/CT were built based on SVM classifier;an ensemble learning model was constructed based on the relative majority voting criterion.The experiments of validity and feasibility were carried out:Delta determination experiment of neighborhood size;experiment of variation coefficient F,cross coefficient CR,weight value w;experiment of ensemble;comparative experiment.The results show that our model has better overall performance,and the accuracy is 99.72%.It has good robustness and scalability.

关 键 词：邻域粗糙集 差分进化 肺部肿瘤 计算机辅助诊断 集成支持向量机 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]