基于深度卷积神经网络的人体外周血白细胞显微图像分类
|
摘要
人体外周血白细胞五分类在医学临床诊断中有重要的作用。本文提出一种基于深度卷积神经网络(CNN)的人体外周血白细胞显微图像五分类方法。首先以ResNet为原型结构设计了一种适用于白细胞显微图像分类的深度卷积神经网络,并提出了一种基于特征集中的新的数据增强的方法来丰富数据集。由于图像的背景对物体识别有很大影响,用图像处理的方法改变同一白细胞的背景,可以生成新的样本。经过数据增强后的样本总量为42 300。最后,针对数据集中五类白细胞样本不均衡问题,在神经网络训练策略中,提出一种改进的批次(batch)随机梯度下降算法(MBGD)。通过将批次随机梯度下降算法每个批次中五类白细胞所占比例设置为1∶1∶1∶1∶1,可以使CNN均衡地获取五类白细胞的特征。实验结果表明,本文所设计的CNN结构、所提出数据增强方法和改进的批次随机梯度下降算法均可提高白细胞图像分类正确率。所提白细胞五分类方法可以达到95.7%的训练正确率。对8 400张白细胞图像进行测试,得到95.0%的平均分类正确率,嗜中性粒细胞、淋巴细胞、单核细胞、嗜酸性粒细胞和嗜碱性粒细胞的分类正确率分别为:92.2%,91.5%,94.6%,93.3%和97.4%。
Human peripheral blood leukocyte five kinds classification plays a vital role in clinical diagnose.In this paper,we propose a method of human peripheral blood leukocyte microscopic image classification based on deep Convolutional Neural Network(CNN).Firstly,a deep CNN architecture based on ResNet is designed and it is applicable to leukocyte microscopic image.Then a new data augmentation method based on feature focus is proposed to enrich our dataset.Considering that the surroundings are important for object recognition,we generate a large number of images by putting a segmented leukocyte in different microscopic image surroundings using image processing.After data augmentation,the total amount of the dataset is 42 300.Finally,aiming at the disproportion of five kinds of leukocyte in dataset,we propose a modified batch stochastic gradient descent(MBGD) to train the CNN model.By setting the ratio of five kinds of leukocytes into 1∶1∶1∶1∶1 in a batch,CNN model can evenly achieve features of five kinds of leukocyte.Experimental results demonstrate that the designed CNN architecture,proposed data augmentation method and modified stochastic gradient descent can all improve the classification accuracy.The proposed method can achieve 95.7% training accuracy.The average testing accuracy of 8 400 leukocyte images is 95.0%.The accuracy of neutrophils,lymphocytes,monocytes,eosinophils and basophils are respectively 92.2%,91.5%,94.6%,93.3% and 97.4%.
Human peripheral blood leukocyte five kinds classification plays a vital role in clinical diagnose.In this paper,we propose a method of human peripheral blood leukocyte microscopic image classification based on deep Convolutional Neural Network(CNN).Firstly,a deep CNN architecture based on ResNet is designed and it is applicable to leukocyte microscopic image.Then a new data augmentation method based on feature focus is proposed to enrich our dataset.Considering that the surroundings are important for object recognition,we generate a large number of images by putting a segmented leukocyte in different microscopic image surroundings using image processing.After data augmentation,the total amount of the dataset is 42 300.Finally,aiming at the disproportion of five kinds of leukocyte in dataset,we propose a modified batch stochastic gradient descent(MBGD) to train the CNN model.By setting the ratio of five kinds of leukocytes into 1∶1∶1∶1∶1 in a batch,CNN model can evenly achieve features of five kinds of leukocyte.Experimental results demonstrate that the designed CNN architecture,proposed data augmentation method and modified stochastic gradient descent can all improve the classification accuracy.The proposed method can achieve 95.7% training accuracy.The average testing accuracy of 8 400 leukocyte images is 95.0%.The accuracy of neutrophils,lymphocytes,monocytes,eosinophils and basophils are respectively 92.2%,91.5%,94.6%,93.3% and 97.4%.
引文
[1]Shirazi Syed H,Umar Arif Lqbal,Naz Saeeda,et al.Efficient leukocyte segmentation and recognition in peripheral blood image[J].Technology and Health Care,2016,24(3):335-347.
[2]Rawat Jyoti,Annapurna Singh,Bhadauria H S,et al.Computer aided diagnostic system for detection of leukemia using microscopic images[J].Procedia Computer Science,2015,70:748-756.
[3]Mohamed Hend,Omar Rowan,Saeed Nermeen,et al.Automated detection of white blood cells cancer diseases[J].In:Deep and Representation Learning(IWDRL),2018 First International Workshop on.IEEE,2018,48-54.
[4]Biji G,Hariharan S.Leukocyte segmentation and classification techniques for leukaemia detection[J].International Journal of Electronics,Electrical and Computational System,2017,6(5):242-251.
[5]Angulo J,Flandrin G.Automated detection of working area of peripheral blood smears using mathematical morphology[J].Analytical cellular pathology,2003,25(1):37-49.
[6]Safuan S N M,Tomari M R.White blood cell(WBC)counting analysis in blood smear images using various color segmentation methods[J].Measurement,2018,116:543-555.
[7]Ananthi V P,Balasubramaniam P.A new thresholding technique based on fuzzy set as an application to leukocyte nucleus segmentation[J].Computer methods and programs in biomedicine,2016,134:165-177.
[8]Jiang K,Liao Q M,Xiong Y.A novel white blood cell segmentation scheme based on feature space clustering[J].Soft Computing,2006,10(1):12-19.
[9]Marzukia N I Che,Mahmoodb Nasrul Humaimi,Razakb MAAbdul.Segmentation of white blood cell nucleus using active contour[J].Jurnal Teknologi,2015,74(6):115-118.
[10]LIU Yue-hua,CAO Fei-long,ZHAO Jian-wei,et al.Segmentation of white blood cells image using adaptive location and iteration[J].IEEE journal of biomedical and health informatics,2017,21(6):1644-1655.
[11]Sabino Daniela Mayumi Ushizima,Costa L F,Rizzatti,EG,et al.Toward leukocyte recognition using morphometry,texture and color[J].In:Biomedical Imaging:Nano to Macro,2004.IEEE International Symposium on,2004,121-124.
[12]PANG Gai,ZHUANG Yang-kai,ZHOU Ping.Automatic leukocytes classification by distance transform,moment invariant,morphological features,gray level co-occurrence matrices and SVM[J].In:International Conference on Information Sciences,Machinery,Materials and Energy(ICISMME2015),2015,1090-1095.
[13]Mishra Sonali,Majhi Banshidhar,Sa Pankaj Kumar,et al.Gray level co-occurrence matrix and random forest based acute lymphoblastic leukemia detection[J].Biomedical Signal Processing and Control,2017,33:272-280.
[14]Gautam A,Singh P,Raman B.Automatic classification of leukocytes using morphological features and na¨1ve Bayes classifier[J].In Region 10 Conference(TENCON)IEEE,2016,1023-1027.
[15]Kratz A,Bengtsson H I,Casey J E.Performance evaluation of the CellaVision DM96 system:WBC differentials by automated digital image analysis supported by an artificial neural network[J].A-merican journal of clinical pathology,2005,124(5):770-781.
[16]ZHONG Ya,ZHANG Jing,Xiao Jun-ji.Automatic detection of leukocytes in leucorrhea based on convolution neural network[J].Chinese Journal of Biomedical Engineering,2018,37(2):163-168.钟亚,张静,肖峻基.于卷积神经网络的白带中白细胞的自动检测[J].中国生物医学工程学报,2018,37(2):163-168.
[17]Zhao J,Zhang M,Zhou Z.Automatic detection and classification of leukocytes using convolutional neural networks[J].Medical&biological engineering&computing,2017,55(8):1287-1301.
[18]Krizhevsky Alex,Sutskever Ilya,Hinton Geoffrey E.Imagenet classification with deep convolutional neural networks[J].In:Advances in neural information processing systems,2012,1097-1105.
[19]Simonyan K,Zisserman A.Very deep convolutional networks for large-scale image recognition[J].arXiv preprint arXiv:1409.1556,2014.
[20]He K,Zhang X,Ren S.Deep residual learning for image recognition.Proceedings of the IEEE conference on computer vision and pattern recognition,2016,770-778.
[21]Szegedy,Christian,Wei liu,Yangqing Jia,et al.Going deeper with convolutions[J].In:Proceedings of the IEEE conference on computer vision and pattern recognition,2015,1-9.
[22]HUANG Gao,LIU Zhuang,Laurens van der Maaten,et al.Densely connected convolutional networks[J].In:CVPR.2017,1(2):4700-4708.
[23]Salamon J,Bello J P.Deep convolutional neural networks and data augmentation for environmental sound classification[J].IEEESignal Processing Letters,2017,24(3):279-283.
[24]HUAI Ting-ting,ZHAO Jian-wei,CAO Fei-long.A classification algorithm for white blood cells based on the synthetic feature and random forest[J].Journal of China University of Metrology,2015,26(4):476-479.怀听听,赵建伟,曹飞龙.基于综合特征和随机森林的白细胞分类算法[J].中国计量学院学报,2015,26(4):476-479.
[25]GLOROT Xavier,BENGIO Yoshua.Understanding the difficulty of training deep feedforward neural networks[J].Proceedings of the thirteenth international conference on artificial intelligence and statistics,2010,249-256.
[26]Saxe Andrew M,Mcclelland James L,Ganguli,Surya.Exact solutions to the nonlinear dynamics of learning in deep linear neural networks[J].arXiv preprint arXiv:1312.6120,2013.
[27]WANG Ya-pin,CAO Yi-png.A Leukocyte image fast scanning based on max-min distance clustering[J].Journal of Innovative Optical Health Sciences,2016,9(6):1650022.
[28]Rumelhart,David E,Geoffrey E.Hinton,Ronald J.Williams.Learning internal representations by error propagation.California Univ San Diego La Jolla Inst for Cognitive Science,1985,No.ICS-8506.
[29]BOTTOU,Léon.Large-scale machine learning with stochastic gradient descent.In:Proceedings of COMPSTAT′2010.Physica-Verlag HD,2010,177-186.
[30]Li M,Zhang T,Chen Y,&Smola,A.J.(2014,August).Efficient mini-batch training for stochastic optimization.In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining(661-670).ACM.
[31]Kline,Douglas M,BERARDI,Victor L.Revisiting squared-error and cross-entropy functions for training neural network classifiers.Neural Computing&Applications,2005,14(4):310-318.
[32]N G,Andrew Y.Feature selection,L 1 vs.L 2 regularization,and rotational invariance.In:Proceedings of the twenty-first international conference on Machine learning.ACM,2004,78.
[33]HE Kai-ming,ZHANG Xiang-yu,REN Shao-qing,et al.Delving deep into rectifiers:Surpassing human-level performance on imagenet classification[J].In:Proceedings of the IEEE international conference on computer vision,2015,1026-1034.
[2]Rawat Jyoti,Annapurna Singh,Bhadauria H S,et al.Computer aided diagnostic system for detection of leukemia using microscopic images[J].Procedia Computer Science,2015,70:748-756.
[3]Mohamed Hend,Omar Rowan,Saeed Nermeen,et al.Automated detection of white blood cells cancer diseases[J].In:Deep and Representation Learning(IWDRL),2018 First International Workshop on.IEEE,2018,48-54.
[4]Biji G,Hariharan S.Leukocyte segmentation and classification techniques for leukaemia detection[J].International Journal of Electronics,Electrical and Computational System,2017,6(5):242-251.
[5]Angulo J,Flandrin G.Automated detection of working area of peripheral blood smears using mathematical morphology[J].Analytical cellular pathology,2003,25(1):37-49.
[6]Safuan S N M,Tomari M R.White blood cell(WBC)counting analysis in blood smear images using various color segmentation methods[J].Measurement,2018,116:543-555.
[7]Ananthi V P,Balasubramaniam P.A new thresholding technique based on fuzzy set as an application to leukocyte nucleus segmentation[J].Computer methods and programs in biomedicine,2016,134:165-177.
[8]Jiang K,Liao Q M,Xiong Y.A novel white blood cell segmentation scheme based on feature space clustering[J].Soft Computing,2006,10(1):12-19.
[9]Marzukia N I Che,Mahmoodb Nasrul Humaimi,Razakb MAAbdul.Segmentation of white blood cell nucleus using active contour[J].Jurnal Teknologi,2015,74(6):115-118.
[10]LIU Yue-hua,CAO Fei-long,ZHAO Jian-wei,et al.Segmentation of white blood cells image using adaptive location and iteration[J].IEEE journal of biomedical and health informatics,2017,21(6):1644-1655.
[11]Sabino Daniela Mayumi Ushizima,Costa L F,Rizzatti,EG,et al.Toward leukocyte recognition using morphometry,texture and color[J].In:Biomedical Imaging:Nano to Macro,2004.IEEE International Symposium on,2004,121-124.
[12]PANG Gai,ZHUANG Yang-kai,ZHOU Ping.Automatic leukocytes classification by distance transform,moment invariant,morphological features,gray level co-occurrence matrices and SVM[J].In:International Conference on Information Sciences,Machinery,Materials and Energy(ICISMME2015),2015,1090-1095.
[13]Mishra Sonali,Majhi Banshidhar,Sa Pankaj Kumar,et al.Gray level co-occurrence matrix and random forest based acute lymphoblastic leukemia detection[J].Biomedical Signal Processing and Control,2017,33:272-280.
[14]Gautam A,Singh P,Raman B.Automatic classification of leukocytes using morphological features and na¨1ve Bayes classifier[J].In Region 10 Conference(TENCON)IEEE,2016,1023-1027.
[15]Kratz A,Bengtsson H I,Casey J E.Performance evaluation of the CellaVision DM96 system:WBC differentials by automated digital image analysis supported by an artificial neural network[J].A-merican journal of clinical pathology,2005,124(5):770-781.
[16]ZHONG Ya,ZHANG Jing,Xiao Jun-ji.Automatic detection of leukocytes in leucorrhea based on convolution neural network[J].Chinese Journal of Biomedical Engineering,2018,37(2):163-168.钟亚,张静,肖峻基.于卷积神经网络的白带中白细胞的自动检测[J].中国生物医学工程学报,2018,37(2):163-168.
[17]Zhao J,Zhang M,Zhou Z.Automatic detection and classification of leukocytes using convolutional neural networks[J].Medical&biological engineering&computing,2017,55(8):1287-1301.
[18]Krizhevsky Alex,Sutskever Ilya,Hinton Geoffrey E.Imagenet classification with deep convolutional neural networks[J].In:Advances in neural information processing systems,2012,1097-1105.
[19]Simonyan K,Zisserman A.Very deep convolutional networks for large-scale image recognition[J].arXiv preprint arXiv:1409.1556,2014.
[20]He K,Zhang X,Ren S.Deep residual learning for image recognition.Proceedings of the IEEE conference on computer vision and pattern recognition,2016,770-778.
[21]Szegedy,Christian,Wei liu,Yangqing Jia,et al.Going deeper with convolutions[J].In:Proceedings of the IEEE conference on computer vision and pattern recognition,2015,1-9.
[22]HUANG Gao,LIU Zhuang,Laurens van der Maaten,et al.Densely connected convolutional networks[J].In:CVPR.2017,1(2):4700-4708.
[23]Salamon J,Bello J P.Deep convolutional neural networks and data augmentation for environmental sound classification[J].IEEESignal Processing Letters,2017,24(3):279-283.
[24]HUAI Ting-ting,ZHAO Jian-wei,CAO Fei-long.A classification algorithm for white blood cells based on the synthetic feature and random forest[J].Journal of China University of Metrology,2015,26(4):476-479.怀听听,赵建伟,曹飞龙.基于综合特征和随机森林的白细胞分类算法[J].中国计量学院学报,2015,26(4):476-479.
[25]GLOROT Xavier,BENGIO Yoshua.Understanding the difficulty of training deep feedforward neural networks[J].Proceedings of the thirteenth international conference on artificial intelligence and statistics,2010,249-256.
[26]Saxe Andrew M,Mcclelland James L,Ganguli,Surya.Exact solutions to the nonlinear dynamics of learning in deep linear neural networks[J].arXiv preprint arXiv:1312.6120,2013.
[27]WANG Ya-pin,CAO Yi-png.A Leukocyte image fast scanning based on max-min distance clustering[J].Journal of Innovative Optical Health Sciences,2016,9(6):1650022.
[28]Rumelhart,David E,Geoffrey E.Hinton,Ronald J.Williams.Learning internal representations by error propagation.California Univ San Diego La Jolla Inst for Cognitive Science,1985,No.ICS-8506.
[29]BOTTOU,Léon.Large-scale machine learning with stochastic gradient descent.In:Proceedings of COMPSTAT′2010.Physica-Verlag HD,2010,177-186.
[30]Li M,Zhang T,Chen Y,&Smola,A.J.(2014,August).Efficient mini-batch training for stochastic optimization.In Proceedings of the 20th ACM SIGKDD international conference on Knowledge discovery and data mining(661-670).ACM.
[31]Kline,Douglas M,BERARDI,Victor L.Revisiting squared-error and cross-entropy functions for training neural network classifiers.Neural Computing&Applications,2005,14(4):310-318.
[32]N G,Andrew Y.Feature selection,L 1 vs.L 2 regularization,and rotational invariance.In:Proceedings of the twenty-first international conference on Machine learning.ACM,2004,78.
[33]HE Kai-ming,ZHANG Xiang-yu,REN Shao-qing,et al.Delving deep into rectifiers:Surpassing human-level performance on imagenet classification[J].In:Proceedings of the IEEE international conference on computer vision,2015,1026-1034.