北京生物医学工程

基于卷积神经网络的眼科光学相干断层成像图像的自动分类

Automatic classification of ophthalmic optical coherence tomographyimages based on the convolution neural network

作者：赵蒙蒙鲁贞贞朱书缘王小兵冯继宏

单位：北京工业大学环境与生命学部(北京100124) ,首都医科大学附属北京同仁医院眼科中心(北京 100730) 通信作者:冯继宏，E-mail： jhfeng@ bjut. edu. cn；王小兵，E-mail: little-bill@ 263. net

关键词：眼科疾病;光学相干断层成像;图像处理;卷积神经网络;图像分类

分类号：R318. 04

出版年·卷·期（页码）：2021·40·6（557-563）

摘要：

目的提出一种基于卷积神经网络(convolutional neural network, CNN)的眼科光学相干断层成像(optical coherence tomography,OCT)图像自动分类方法,实现对视网膜OCT图像的自动分类，缓解人工诊断依赖医生的临床经验、费时费力等问题。方法基于公开的数据集2014_BOE_Srinivasan构建了 2个样本数据集。其中样本数据集一为仅对数据集中的图像进行预处理后裁剪，样本数据集二为对取出测试集后剩余图像的裁剪过程中引入随机平移和水平翻转技术对图像进行扩充,并划分为训练集和验证集。搭建基于CNN的视网膜OCT图像分类网络，并分别使用两个数据集训练网络得到分类模型。最后使用独立的测试集对模型进行测试，并通过输出混淆矩阵查看模型对3种类别图像的分类情况。结果通过混淆矩阵计算得出，使用扩充后的图像训练的分类模型的准确度为93. 43%,灵敏度为 91.38%,特异度为95. 88%。结论提出的基于CNN的视网膜OCT图像自动分类方法可以对老年性黄斑变性、糖尿病性黄斑水肿和正常3种类别的视网膜OCT图像进行分类。同时,数据扩充有助于提高分类算法的性能。

Objective To propose an automatic classification method of ophthalmic optical coherence tomography ( OCT) images based on the convolutional neural network ( CNN ) , and alleviate the problems of artificial diagnosis relying on the clinical experience of ophthalmologists. Methods The two sample data sets were constructed based on the OCT data set of 2014 BOE Srinivasan. Among them, the sample data set 1 was to crop the images in the data set after preprocessing, the sample data set 2 was to augment the images by introducing the random translation and horizontal flipping technology in the images cropping of remaining images after taking out the test set, and it was divided into training set and validation set. Then a classification network based on the CNN was constructed and respectively trained by the two sample data sets to acquire the classification models. Finally, the model was tested by using the test set,and the accuracy of the model for the three types of retinal OCT images were calculated through the output confusion matrix. Results The accuracy, the sensitivity and the specificity of the model trained by augmented images were 93. 43%, 91. 38% and 95. 88%, respectively. Conclusions The proposed automatic classification method of retinal OCT images based on the CNN can classify the retinal OCT images of age-related macular degeneration, diabetic macular edema and normal. Meanwhile, data augmentation can improve the performance of classification algorithm.

参考文献：

[1 ] Coleman HR, Chan CC, Fems III FL, et al. Age-related macular degeneration [ J]. The Lancet,2008,372(9652) : 1835-1845.

[2 ] Jager RD, Mieler WF, Miller JW. Medical progress: age-related macular degeneration [ J]. New England Journal of Medicine, 2008,358(24)：2606-2617.

[3 ] Ciulla TA, Amador AG, Zinman B. Diabetic retinopathy and diabetic macular edema-pathophysiology, screening, and novel therapies [ J]. Diabetes Care,2003 ,26(9) :2653-2664.

[4 ] Lang GE. Diabetic macular edema [ J ]. Ophthalmologica, 2012, 227(Suppl l)：21-29.

[5 ] Bressler NM. Age-related macular degeneration is the leading cause of blindness [J]. JAMA,2004,291( 15) : 1900-1901.

[6 ] Farsiu S,Chiu SJ,O' connell RV,et al. Quantitative classification of eyes with and without intermediate age-related macular degeneration using optical coherence tomography [ J ]. Ophthalmology ,2014,121 ( 1 )： 162-172.

[7 ] Huang D, Swanson EA, Lin CP, et al. Optical coherence tomography [ J]. Science, 1991,254(5035) ： 1178-1181.

[8 ] Krizhevsky A, Sutskever I, Hinton GE. ImageNet Classification with Deep Convolutional Neural Networks [ J]. Communications of the ACM,2017,60(6) ：84-90.

[9 ] Litjens G,Kooi T,Bejnordi BE,et al. A survey on deep learning in medical image analysis [ J]. Medical Image Analysis, 2017, 42：60-88.

[10] Lee CS, Tyring AJ, Deruyter NP, et al. Deep-learning based, automated segmentation of macular edema in optical coherence tomography [ J ]. Biomedical Optics Express, 2017, 8(7)： 3440-3448.

[11 ] Karri SPK, Chakraborty D, Chatterjee J. Transfer learning based classification of optical coherence tomography images with diabetic macular edema and dry age-related macular degeneration [J]. Biomedical Optics Express,2017,8(2) ：579.

[12]?Awais M, Muller H, Tang TB, et al. Classification of SD - OCT images using a Deep learning approach [ C ]// 2017 IEEE International Conference on Signal and Image Processing Applications. Kuching, Malaysia: IEEE Press ,2017:489-492.

[13]?Srinivasan PP, Kim LA, Mettu PS, et al. Fully automated detection of diabetic macular edema and dry age-related macular degeneration from optical coherence tomography images [ J ]. Biomedical Optics Express,2014,5( 10) ：3568-3577.

[14]?张茂鑫，李国春.基于HDF5文件格式的MERSI影像数据提取的研究与实现[J].现代农业科学,2009(3) ： 189-191,222. Zhang MX, Li GC. The study and implementation of extraction MERSI image data based on the file format of HDF[ J]. Modem Agricultural Science,2009 (3) ：189-191,222.

[15]?李飞腾.卷积神经网络及其应用[D].大连:大连理工大学， 2014：20-25.

Li FT. Convolutional neural network and its applications [ D ]. Dalian: Dalian University of Technology ,2014： 20—25.

[16]?周飞燕，金林鹏，董军.卷积神经网络研究综述[J].计算机学报,2017,40(6) ： 1229-1251.

Zhou FY, Jin LP, Dong J. Review of convolutional neural networks [ J]. Chinese Journal of Computers, 2017,40 ( 6)： 1229-1251.

[17]?王双印，滕国文.卷积神经网络中ReLU激活函数优化设计 [J].信息通信,2018(1) ：42-43,

Wang SY, Teng GW. Optimal design of ReLU activation function in convolutional neural networks [ J ]. Information and Communication,2018 ( 1) ：42-43.

[18]?袁铭阳，周长胜,黄宏博，等.卷积神经网络池化方法综述 [J].软件工程与应用,2020,9(5):360-372.

Yuan MY, Zhou CS, Huang HB, et al. Survey on convolutional neural networks pooling methods [ J]. Software Engineering and Applications, 2020,9(5) : 360-372.

[19]?程俊华，曾国辉,鲁敦科，等.基于Dropout的改进卷积神经网络模型平均方法[J].计算机应用,2019,39(6):1601-1606.

Cheng JH,Zeng GH,Lu DK,et al. Improved convolutional neural network model averaging method based on dropout [ J] . Journal of Computer Applications,2019,39(6) : 1601-1606.

[20]?李理，应三丛.基于FPGA的卷积神经网络Softmax层实现 [J].现代计算机(专业版),2017 ( 26) ：21- 24.

Li L, Ying SC. Softmax layer implementation of neural network based on FPGA [ J] . Modem Computer,2017 ( 26) ：21—24.

[21 ] Chen MH, He JT, Jia WY, et al. Segmentation of diabetic macular edema for retinal OCT images [ C ]//Conference on Optics in Health Care and Biomedical Optics VIII. Beijing, China: SPIE, 2018,10820：108200X.

服务与反馈：

【文章下载】【加入收藏】

提示：您还未登录，请登录！点此登录