基于粒子群算法的医学图像分类算法研究
|
摘要
针对医学图像具有较大的相似性和交叉性,易造成归属类别混乱的问题,提出了一种基于粒子群算法的医学图像分类方法。该方法使用形态学滤波和阈值法进行预处理;使用SIFT特征描述子来提取图像的局部特征,并使用聚类的方法得到SIFT特征的"视觉词汇";使用粒子群算法选出一些列多样性和精度更高的SVM、KNN和AdaBoost分类器对特征进行分类。对15种不同类型的医学图像进行分类的结果表明,所提出的方法取得了94.72%的分类精度,且相比于单个分类器的方法具有较大的性能提升。
In view of the great similarity and cross-over of medical images and the confusion of attribution categories,a medical image classification method based on particle swarm optimization is proposed. The method uses morphological filtering and thresholding method to preprocess;SIFT feature descriptors are used to extract the local features of the image,and SIFT feature"visual vocabulary"is obtained by using clustering method;a number of columns are selected by using the particle swarm algorithm. The SVM,KNN,and AdaBoost classifiers,which are more accurate and more accurate,classify features. The results of the classification of 15 different types of medical images show that the proposed method achieves a classification accuracy of 94.72%, and has a greater performance improvement compared to the method of a single classifier.
In view of the great similarity and cross-over of medical images and the confusion of attribution categories,a medical image classification method based on particle swarm optimization is proposed. The method uses morphological filtering and thresholding method to preprocess;SIFT feature descriptors are used to extract the local features of the image,and SIFT feature"visual vocabulary"is obtained by using clustering method;a number of columns are selected by using the particle swarm algorithm. The SVM,KNN,and AdaBoost classifiers,which are more accurate and more accurate,classify features. The results of the classification of 15 different types of medical images show that the proposed method achieves a classification accuracy of 94.72%, and has a greater performance improvement compared to the method of a single classifier.
引文
[1]周慧,张尤赛,龚淼.基于RBF神经网络的医学图像分类算法研究[J].电子设计工程,2017,25(3):113-116.
[2]岳丽娟,姜英姿.基于自适应混沌粒子群和支持向量机的医学图像分类[J].医疗装备,2017,30(3):23-26.
[3]孟志伟,刘惠义,陈霜霜.基于RBM-KNN的脑部磁共振图像分类[J].信息技术,2017(4):169-173.
[4]赵梦旸,那彦,李思彤.基于直觉模糊推理的医学图像融合方法研究[J].电子科技,2012,25(3):48-50.
[5] Yan S,Xu X,Xu D,et al. Image classificationwith densely sampled image windows andgeneralized adaptive multiple kernel learning[J].IEEE Transactions on Cybernetics,2017,45(3):381-390.
[6] Zhu X,Li X,Zhang S.Block-row sparse multiviewmultilabel learning for image classification[J].IEEE Transactions on Cybernetics,2017,46(2):450-461.
[7] Wang Q,Lin J,Yuan Y. Salient band selection forhyperspectral image classification via manifoldranking[J]. IEEE Transactions on Neural Networks&Learning Systems,2017,27(6):1279-1289.
[8] Li P,Wang Q,Hui Z,et al.Local log-euclideanmultivariate gaussian descriptor and its applicationto image classification[J]. IEEE Transactions onPattern Anal Mach Intell,2017,39(4):803-817.
[9]刘海玲,裴连群.基于遗传算法的医学图像配准方法改进[J].自动化与仪器仪表,2016(9):218-220.
[10]韦明祥,陈俊.基于C-V模型的医学图像分割方法[J].电子科技,2012,25(5):101-104.
[11]陈贞,邢笑雪.基于非下采样剪切波变换的医学图像融合算法[J].沈阳工业大学学报,2015,37(2):194-199.
[12]Wan S,Lee H C,Huang X,et al. Integrated localbinary pattern texture features for classification ofbreast tissue imaged by optical coherencemicroscopy[J].Medical Image Analysis,2017(38):104-116.
[13]Manivannan S,Cobb C,Burgess S,et al. Subcategory classifiers for multiple-instance learningand its application to retinal nerve fiber layervisibility classification[J].IEEE Transactions onMedical Imaging,2017,36(5):1140-1150.
[14]崔宝侠,田佳,段勇,等.基于图论分割的肺部CT图像的三维重建[J].沈阳工业大学学报,2015,37(6):667-672.
[15]Lee S,Charon N,Charlier B,et al.Atlas-basedshape analysis and classification of retinal opticalcoherence tomography images using the functionalshape(fshape)framework[J].Medical Image Analy-sis,2017(35):570-581.
[16]Zimmer V A,Glocker B,Hahner N,et al. Learn-ing and combining image neighborhoods using ran-dom forests for neonatal brain disease classification[J].Medical Image Analysis,2017,42(6):189-198.
[17]Wang S S,Wang Q. Lung cancer image fine-grained classification based on wavelet momentfused with LBP[J].Journal of Northeast NormalUniversity,2017(7):533-548.
[2]岳丽娟,姜英姿.基于自适应混沌粒子群和支持向量机的医学图像分类[J].医疗装备,2017,30(3):23-26.
[3]孟志伟,刘惠义,陈霜霜.基于RBM-KNN的脑部磁共振图像分类[J].信息技术,2017(4):169-173.
[4]赵梦旸,那彦,李思彤.基于直觉模糊推理的医学图像融合方法研究[J].电子科技,2012,25(3):48-50.
[5] Yan S,Xu X,Xu D,et al. Image classificationwith densely sampled image windows andgeneralized adaptive multiple kernel learning[J].IEEE Transactions on Cybernetics,2017,45(3):381-390.
[6] Zhu X,Li X,Zhang S.Block-row sparse multiviewmultilabel learning for image classification[J].IEEE Transactions on Cybernetics,2017,46(2):450-461.
[7] Wang Q,Lin J,Yuan Y. Salient band selection forhyperspectral image classification via manifoldranking[J]. IEEE Transactions on Neural Networks&Learning Systems,2017,27(6):1279-1289.
[8] Li P,Wang Q,Hui Z,et al.Local log-euclideanmultivariate gaussian descriptor and its applicationto image classification[J]. IEEE Transactions onPattern Anal Mach Intell,2017,39(4):803-817.
[9]刘海玲,裴连群.基于遗传算法的医学图像配准方法改进[J].自动化与仪器仪表,2016(9):218-220.
[10]韦明祥,陈俊.基于C-V模型的医学图像分割方法[J].电子科技,2012,25(5):101-104.
[11]陈贞,邢笑雪.基于非下采样剪切波变换的医学图像融合算法[J].沈阳工业大学学报,2015,37(2):194-199.
[12]Wan S,Lee H C,Huang X,et al. Integrated localbinary pattern texture features for classification ofbreast tissue imaged by optical coherencemicroscopy[J].Medical Image Analysis,2017(38):104-116.
[13]Manivannan S,Cobb C,Burgess S,et al. Subcategory classifiers for multiple-instance learningand its application to retinal nerve fiber layervisibility classification[J].IEEE Transactions onMedical Imaging,2017,36(5):1140-1150.
[14]崔宝侠,田佳,段勇,等.基于图论分割的肺部CT图像的三维重建[J].沈阳工业大学学报,2015,37(6):667-672.
[15]Lee S,Charon N,Charlier B,et al.Atlas-basedshape analysis and classification of retinal opticalcoherence tomography images using the functionalshape(fshape)framework[J].Medical Image Analy-sis,2017(35):570-581.
[16]Zimmer V A,Glocker B,Hahner N,et al. Learn-ing and combining image neighborhoods using ran-dom forests for neonatal brain disease classification[J].Medical Image Analysis,2017,42(6):189-198.
[17]Wang S S,Wang Q. Lung cancer image fine-grained classification based on wavelet momentfused with LBP[J].Journal of Northeast NormalUniversity,2017(7):533-548.