基于集成VPRS-RUGGA支持向量机的多模态肺部肿瘤计算机辅助诊断模型
|
摘要
针对计算机辅助诊断模型优化过程中稳定性差和早熟问题,提出基于集成VPRS-RUGGA-支持向量机的肺部肿瘤计算机辅助诊断模型。首先,引入变精度粗糙集构造属性依赖度,结合属性约简长度和惩罚函数的加权和构造适应度函数框架;其次,采用无回放余数随机选择法、均匀交叉和高斯变异算子进行遗传操作;然后,在CT、PET和PET/CT样本空间中提取肺部肿瘤ROI区域特征,构造不同的特征空间,运用VPRS-RUGGA-支持向量机模型约简和分类识别;最后,在不同的样本空间中构造支持向量机(SVM)个体分类器,采用相对多数投票法输出集成结论。实验结果表明,集成VPRS-RUGGA-SVM模型可以有效的提高泛化性能和稳定性,VPRS-RUGGA-SVM模型可有效改善早熟问题,提高模型的分类性能。
Aiming at the problem of poor stability and premature in computer aided diagnosis model optimization procedure, to propose a computer aided diagnosis model of lung tumor based on ensemble VPRS-RUGGA-Support vector machine(SVM).Firstly,a fitness function framework was constructed based on weighted sum of three objective functions of variable precision rough set attribute dependency, attribute reduction length and penalty function;Secondly, remainder stochastic sampling with replacement,uniform crossover and Gaussian mutation operator were used to carry out genetic manipulation;Then,ROI region features were extracted of lung tumor in three sample spaces of CT, PET and PET/CT to construct different feature spaces, and using the VPRS-RUGGA-SVM model to reduce and classify.Finally, the SVM individual classifier was constructed in different sample spaces, and the relative majority voting method was used to output the ensemble conclusion. The experimental results show that the ensemble VPRS-RUGGA-SVM model can effectively improve the generalization performance and stability. VPRS-RUGGA-SVM model can relieve premature and improve the classification performance.
Aiming at the problem of poor stability and premature in computer aided diagnosis model optimization procedure, to propose a computer aided diagnosis model of lung tumor based on ensemble VPRS-RUGGA-Support vector machine(SVM).Firstly,a fitness function framework was constructed based on weighted sum of three objective functions of variable precision rough set attribute dependency, attribute reduction length and penalty function;Secondly, remainder stochastic sampling with replacement,uniform crossover and Gaussian mutation operator were used to carry out genetic manipulation;Then,ROI region features were extracted of lung tumor in three sample spaces of CT, PET and PET/CT to construct different feature spaces, and using the VPRS-RUGGA-SVM model to reduce and classify.Finally, the SVM individual classifier was constructed in different sample spaces, and the relative majority voting method was used to output the ensemble conclusion. The experimental results show that the ensemble VPRS-RUGGA-SVM model can effectively improve the generalization performance and stability. VPRS-RUGGA-SVM model can relieve premature and improve the classification performance.
引文
[1]Novello S,Stabile L P,Siegfried J M.5-Gender-related differences in lung cancer[J].Iaslc Thoracic Oncology (Second Edition),2018:30-45.
[2]魏兴瑜,陆惠玲,周涛.两模态PET/CT图像融合研究进展[J].重庆医学,2015,44(14):1979-1982.
[3]Sachdeva J,Kumar V,Gupta I,et al.A package- SFERCB-Segmentation,feature extraction,reduction and classification analysis by both SVM and ANN for brain tumors[J].Applied Soft Computing,2016,47:151-167.
[4]Rendon-Gonzalez E,Ponomaryov V.Autom- atic lung nodule segmentation and classification in CT images based on SVM[C]//2016 9th International Kharkiv Symposium on Physics and Engineering of Microwaves.IEEE,2016.
[5]Wang Ziqi,Zhang Xuefeng,Yang Wanyu.Rule induction of breast cancer medical diagnose based on combination of rough sets,artificial neutral network and genetic algorithm [C]//2016 Chinese Control And Decision Conference.IEEE,2016.
[6]尚丹.基于参数优化的SVM分类器在肺癌早期诊断中的应用[D].郑州大学,2014.
[7]王晓丹,高晓峰,姚旭等.SVM集成研究与应用[J].空军工程大学学报(自然科学版),2012,13(2):84-89.
[8]Chen Degang,Yang Yanyan,Dong Ze.An incremental algorithm for attribute reduction with variableprecision rough sets[J].Applied Soft Computing,2016,45:129-149.
[9]熊方,张贤勇.变精度粗糙集的区域属性约简及其结构启发算法[J].计算机应用,2016,36(11):2954-2957.
[10]李煜,裴宇航,刘景森.融合均匀变异与高斯变异的蝙蝠优化算法[J].控制与决策,2017,32(10):1775-1781.
[11]倪志伟,张琛,倪丽萍.基于萤火虫群优化算法的选择性集成雾霾天气预测方法[J].模式识别与人工智能,2016,29(2):143-153.
[12]姜枫,顾庆,郝慧珍,等.基于内容的图像分割方法综述[J].软件学报,2017,28(1):160-183.
[13]袁小翠,吴禄慎,陈华伟.基于Otsu方法的钢轨图像分割[J].光学精密工程,2016,24(7):1772-1781.
[14]吴翠颖,周涛,陆惠玲,等.基于集成SVM的肺部肿瘤PET/ CT三模态计算机辅助诊断方法[J].生物医学工程研究,2017,36(3):207-212.
[2]魏兴瑜,陆惠玲,周涛.两模态PET/CT图像融合研究进展[J].重庆医学,2015,44(14):1979-1982.
[3]Sachdeva J,Kumar V,Gupta I,et al.A package- SFERCB-Segmentation,feature extraction,reduction and classification analysis by both SVM and ANN for brain tumors[J].Applied Soft Computing,2016,47:151-167.
[4]Rendon-Gonzalez E,Ponomaryov V.Autom- atic lung nodule segmentation and classification in CT images based on SVM[C]//2016 9th International Kharkiv Symposium on Physics and Engineering of Microwaves.IEEE,2016.
[5]Wang Ziqi,Zhang Xuefeng,Yang Wanyu.Rule induction of breast cancer medical diagnose based on combination of rough sets,artificial neutral network and genetic algorithm [C]//2016 Chinese Control And Decision Conference.IEEE,2016.
[6]尚丹.基于参数优化的SVM分类器在肺癌早期诊断中的应用[D].郑州大学,2014.
[7]王晓丹,高晓峰,姚旭等.SVM集成研究与应用[J].空军工程大学学报(自然科学版),2012,13(2):84-89.
[8]Chen Degang,Yang Yanyan,Dong Ze.An incremental algorithm for attribute reduction with variableprecision rough sets[J].Applied Soft Computing,2016,45:129-149.
[9]熊方,张贤勇.变精度粗糙集的区域属性约简及其结构启发算法[J].计算机应用,2016,36(11):2954-2957.
[10]李煜,裴宇航,刘景森.融合均匀变异与高斯变异的蝙蝠优化算法[J].控制与决策,2017,32(10):1775-1781.
[11]倪志伟,张琛,倪丽萍.基于萤火虫群优化算法的选择性集成雾霾天气预测方法[J].模式识别与人工智能,2016,29(2):143-153.
[12]姜枫,顾庆,郝慧珍,等.基于内容的图像分割方法综述[J].软件学报,2017,28(1):160-183.
[13]袁小翠,吴禄慎,陈华伟.基于Otsu方法的钢轨图像分割[J].光学精密工程,2016,24(7):1772-1781.
[14]吴翠颖,周涛,陆惠玲,等.基于集成SVM的肺部肿瘤PET/ CT三模态计算机辅助诊断方法[J].生物医学工程研究,2017,36(3):207-212.