Microsoft Word 赵荣昌.doc

Size: px

Start display at page:

Download "Microsoft Word 赵荣昌.doc"

异柯尹
5 years ago
Views:

1 第 29 卷第 6 期计算机辅助设计与图形学学报 Vol. 29 No 年 6 月 Journal of Computer-Aided Design & Computer Graphics Jun 聚合多通道特征的青光眼自动检测赵荣昌 1,2), 陈再良 1,2), 段宣初 2,4), 陈奇林 1,2), 刘可 2,4) 2,3)*, 朱承璋 1) ( 中南大学信息科学与工程学院长沙 ) 2) ( 中南大学眼科医学影像处理中心长沙 ) 3) ( 中南大学文学与新闻传播学院长沙 ) 4) ( 中南大学湘雅二医院眼科长沙 ) (anandawork@126.com) :.,,.,, Gabor ;,,, ; 2 (RIM-ONE r2 Drishti_GS), ,,. : 计算机辅助检测 ; 青光眼 ; 形态分析 ; 随机森林 ; 多通道特征聚合 :TP Automated Glaucoma Detection Based on Multi-channel Features from Color Fundus Images Zhao Rongchang 1,2), Chen Zailiang 1,2), Duan Xuanchu 2,4), Chen Qilin 1,2), Liu Ke 2,4), and Zhu Chengzhang 2,3)* 1) (School of Information Science and Engineering, Central South University, Changsha ) 2) (Center for Ophthalmic Image Analysis, Central South University, Changsha ) 3) (The College of Literature and Journalism, Central South University, Changsha ) 4) (Ophthalmology Department of the Second Xiangya Hospital, Central South University, Changsha ) Abstract: Automatic analysis of fundus images is the foundation of computer-aided glaucoma screening and diagnosis. To improve the accuracy of glaucoma screening, a novel glaucoma detection method is proposed based on the multi-channel feature aggregation in fundus images. Firstly, multi-channel features are computed based on color distribution, multi-scale Gabor filters and oriented gradient histogram, and they describe the tiny changes in the morphology and structure of the optic disc. Secondly, a random forest classifier is developed to detect glaucoma based on the multi-channel features and ensemble learning technology. Finally, the glaucoma detection algorithm is tested on two challenging glaucoma datasetss and obtains values of the area under the curve with and , respectively. The experimental results show that the proposed method can improve sensitivity and specificity simultaneously. : ; : 基金项目 : ( , ). 赵荣昌 (1982 ),,,, CCF, ; 陈再良 (1972 ),,,,, CCF, ; 段宣初 (1968 ),,,,, ; 陈奇林 (1992 ),,, ; 刘可 (1981 ),,,, ; 朱承璋 (1978 ),,,, CCF,.

2 第 6 期, 等 : 999 Key words: computer-aided diagnosis; glaucoma; morphological analysis; random forest; multi-channel features aggregation.,,, 10%, 30%. 2020, 630 [1].,,.,,.,,.,,.,,,,,.,, 80%, ;,,.,,.. ( ),.,,.,.,,.,.,,.,, (),,,,.,.,,,.,, [2].,,.,,,.,,.,,,.,,. 1 计算机辅助青光眼诊断技术, 2 : 1),,,,,. 2)

1000 计算机辅助设计与图形学学报第 29 卷,,. 2. 1.1 青光眼几何参数测量, ( 1 ).,,.,,,,.,.,, [3] [4] [5] [6] [7].

2011, Acharya [10],,,, ; 2014, Yadav [11],,,, ; 2015, Akram [12] ( )( ),

3 1000 计算机辅助设计与图形学学报第 29 卷,, 青光眼几何参数测量, ( 1 ).,,.,,,,.,.,, [3] [4] [5] [6] [7]. a. b. 1,,,.,,,.,,.,,, ;,.,,. 1.2 青光眼自动检测方法,,,,.,,,.,.,,,,.,, [8-13].,. 2003, Miglior [8],, ; 2009, Nayak [9], ; 2010, Bock [2], B,,. 2011, Acharya [10],,,, ; 2014, Yadav [11],,,, ; 2015, Akram [12] ( )( ), ;

4 第 6 期, 等 : 1001 [13],,, ;, Chen [14] (convolutional neural networks, CNN), C-CNN,.,,,,.,,,. [15],,,, ( 2 ). 2 2 彩色眼底图多通道特征,,.,,,.,,,,., Gabor., ( 3 ),., ; Gabor, ; (histogram of oriented gradient, HOG).,. 2.1 多尺度 Gabor 能量 Gabor,,,. 3a Gabor. Gabor Gabor,. Gabor a. Gabor b. Gabor c. HOG d. RGB 3

5 1002 计算机辅助设计与图形学学报第 29 卷 Gabor.,, ;, Gabor,,. 3b Gabor 4,,. Gabor 2 2 x0y (, ) e σ x F x y cos λ ; x0 xcos ysin y0 xsin ycos, σ Gabor, θ, λ σ. l θ Gabor l Gabor, Exy (, ;, l) Ixyl (, ; ) Fxy (, ; ). 2.2 HOG 特征 2005,, Dalal [16] HOG. HOG,,,.,., HOG. HOG 36, 36. [17], HOG 11 HOG, 11. 3c 11 HOG,. 2.3 图像外观表征,,.,. [3],,,,.,,.,,. RGB 基于随机森林的青光眼自动检测 (random forest, RF),., RF,.,,,., RF. 3.1 RF 模型的建立和训练 RF {h(x, ξ k ), k = 1,2,,K},, {ξ k }, K RF. X,,., RF..,,.,.,,. Gini,. RF., RF 2,

6 第 6 期, 等 : RF. 1) ;,.. 2),., RF,. 4 RF. j, d h(, v j ): 0,1, v ; Ω ; ξ j j, E ˆ arg max ES (, ), j T S j j,. j, h. 3.2 青光眼自动检测算法, RF. 4,, RF ;, ;,,., j RF,. t, p t (y v), RF T 1 py ( v) p( yv ) (1) T t 1 yˆ arg max p( y v ), y 0,1 (2). ={X, Y}, I.. I. Step1.. T, h,. Step2.. For t = 1 to T If Ω(t), Else F(t) If F(t) > h(t). Else if F(t) < h(t) End End End Step3.. RF = { 1, 2,, T}. Step4.. For t = 1 to T, End Step5. (1)(2),. t 4 RF

7 1004 计算机辅助设计与图形学学报第 29 卷 4 实验结果与分析 RIM-ONE r2 Drishti-GS 2. RIM-ONE r2 455, 255, 200. Drishti-GS 101, 4, 70, 31. 2,.,, 3 19, 3 RGB 3 Gabor , RF 800, Gini. 19 RF, RF. RF,,,, 评价方法 (receiver operating characteristic curve, ROC) (the area under the curve, AUC). ROC. (true positive rate, TPR),. (true positive, TP)(false positive, FP)(true negative, TN)(false negative, FN) 4, TPR=TP/ (TP+FN)., TNR=TN/ (FP+TN). 4.2 实验结果,,. 10, 10, ; 9, 1,. 5 ROC.,,. 1. Drishti-GS 101, 70, Drishti-GS,. 2., AUC. Drishti-GS, 101,,., 5

8 第 6 期, 等 : 1005 表 1 本文算法在 RIM-ONE r2 数据集上的测试结果 AUC RIM-ONE r Drishti-GS 表 2 本文算法与已有算法的比较 AUC [2] [13] [14] ,,,. 4.3 与单通道特征分类的比较, Gabor HOG, 3. RIM-ONE r2, 3 3 AUC.,,,.,. 5 结语表 3 3 种特征在 2 种数据集上的 AUC 值 RIM-ONE r2 Drishti-GS Gabor HOG ,,,.,,, RF,.,,,.,. 参考文献 (References): [1] Yang Diya, Wang Ningli. The second grade national prize for science and technology progress the establishment and application of new prevention and treatment technology of primary open angle glaucoma[j]. Journal of Capital Medical University, 2015, 36(1): 1-7(in Chinese) (, [J]., 2015, 36(1): 1-7) [2] Bock R, Meier J, Nyul L G, et al. Glaucoma risk index: automated glaucoma detection from color fundus images[j]. Medical Image Analysis, 2010, 14(3): [3] Yu T, Ma Y D, Li W. Automatic localization and segmentation of optic disc in fundus image using morphology and level set[c] //Proceedings of the 9th International Symposium on Medical Information and Communication Technology. Los Alamitos: IEEE Computer Society Press, 2015: [4] Cheng J, Liu J, Xu Y W, et al. Superpixel classification based optic disc and optic cup segmentation for glaucoma screening[j]. IEEE Transactions on Medical Imaging, 2013, 32(6): [5] Michael D, Hancox O D. Optic disc size, an important consideration in the glaucoma evaluation[j]. Clinical Eye and Vision Care, 1999, 11(2): [6] Damms T, Dannheim F. Sensitivity and specificity of optic disc parameters in chronic glaucoma[j]. Investigative Ophthalmology & Visual Science, 1993, 34(7): [7] Wang Yuping, Chen Qiang, Lu Shengtao, et al. Classification based quantitative assessment of cup-to-disk ratio in spectral domain optical coherence tomography images[j]. Journal of Computer-Aided Design & Computer Graphics, 2015, 27(8): (in Chinese) (,,,. [J]., 2015, 27(8): ) [8] Miglior S, Guareschi M, Albe E, et al. Detection of glaucomatous visual field changes using the moorfields regression analysis of the Heidelberg retina tomograph[j]. American Journal of Ophthalmology, 2003, 136(1): [9] Nayak J, Rajendra A U, Bhat P S, et al. Automated diagnosis of glaucoma using digital fundus images[j]. Journal of Medical Systems, 2009, 33(5):

9 1006 计算机辅助设计与图形学学报第 29 卷 [10] Acharya U R, Dua S, Du X, et al. Automated diagnosis of glaucoma using texture and higher order spectra features[j]. IEEE Transactions on Information Technology in Biomedicine, 2011, 15(3): [11] Yadav D, Partha Sarathi M, Dutta M K. Classification of glaucoma based on texture features using neural networks[c] //Proceedings of the 7th International Conference on Contemporary Computing. Los Alamitos: IEEE Computer Society Press, 2014: [12] Akram M U, Tariq A, Khalid S, et al. Glaucoma detection using novel optic disc localization, hybrid feature set and classification techniques[j]. Australasian Physical & Engineering Science in Medicine, 2015, 38(4): [13] Cheng J, Yin F, Wong D W K, et al. Sparse dissimilarity- constrained coding for glaucoma screening[j]. IEEE Transactions on Biomedical Engineering, 2015, 62(5): [14] Chen X Y, Xu Y W, Yan S C, et al. Automatic feature learning for glaucoma detection based on deep learning[m] //Lecture Notes in Computer Science. Heldelberg: Springer, 2015, 9351: [15] Dollar P, Appel R, Belongie S, et al. Fast feature pyramids for object detection[j]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2014, 36(8): [16] Dalal N, Triggs B. Histograms of oriented gradients for human detection[c] //Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Los Alamitos: IEEE Computer Society Press, 2005: [17] Felzenszwalb P, McAllester D, Ramanan D. A discriminatively trained, multiscale, deformable part model[c] //Proceedings of IEEE Computer Society Conference on Computer Vision and Pattern Recognition. Los Alamitos: IEEE Computer Society Press, 2008: 1-8

第 2 期王向东等 : 一种运动轨迹引导下的举重视频关键姿态提取方法 257 竞技体育比赛越来越激烈, 为了提高体育训练的效率, 有必要在体育训练中引入科学定量的方法许多

第 2 期王向东等 : 一种运动轨迹引导下的举重视频关键姿态提取方法 257 竞技体育比赛越来越激烈, 为了提高体育训练的效率, 有必要在体育训练中引入科学定量的方法许多 2014 年 4 月图学学报 April 2014 第 35 卷第 2 期 JOURNAL OF GRAPHICS Vol.35 No.2 一种运动轨迹引导下的举重视频关键姿态提取方法王向东 1, 张静文 2, 毋立芳 2, 徐文泉 (1. 国家体育总局体育科学研究所, 北京 100061;2. 北京