第 9 卷第 6 期 计算机辅助设计与图形学学报 Vol. 9 No.6 017 年 6 月 Journal of Computer-Aided Design & Computer Graphics Jun. 017 青光眼视神经头参数与视网膜神经纤维层的相关性分析 徐军 1), 陈强 1,)* 1,3), 牛四杰 1) ( 南京理工大学计算机科学与工程学院南京 10094) ) ( 闽江学院福建省信息处理与智能控制重点实验室福州 35011) 3) ( 济南大学信息科学与工程学院济南 500) (chenqiang@njust.edu.cn) : 青光眼是一种以视神经萎缩和视野缺损为共同特征的视网膜疾病, 是导致人类失明的第二大视网膜疾病. 青光眼的早期症状不明显, 因此对早期青光眼的筛选和诊断将会阻止青光眼的进一步发展. 文中提出一种评估青光眼发病机制的算法, 首先利用随机森林分割视网膜神经纤维层, 然后利用块搜索算法分割视盘与视杯, 最后分析两者相关性. 实验结果表明, 视网膜神经纤维层与垂直杯盘比, 视杯面积以及沿盘面积比的相关性大小为 0.64, 0.6 和 0.54, 验证了在诊断青光眼方面计算视网膜神经纤维层厚度与杯盘比大小是密切相关和互补的, 对研究青光眼的发展趋势具有重要意义. : 频域光学相干断层技术 ; 青光眼 ; 视网膜神经纤维层 ; 杯盘比 ; 随机森林 ; 支持向量机 ; 相关性分析 :TP391.41 Correlation between Optic Nerve Head Parameters and Retinal Nerve Fiber Layer in Glaucoma Xu Jun 1), Chen Qiang 1,)*, and Niu Sijie 1,3) 1) (School of Computer Science and Engineering, Nanjing University of Science and Technology, Nanjing 10094) ) (Fujian Provincial Key Laboratory of Information Processing and Intelligent Control, Minjiang University, Fuzhou 35011) 3) (School of Information Science and Engineering, University of Ji nan, Ji nan 500) Abstract: Glaucoma is a kind of retinal diseases with the feature of optic nerve atrophy and visual field defect, which is the second foremost reason of blindness. Early glaucoma is not obvious, so the screening and diagnosis of early glaucoma will prevent further development of glaucoma. In this paper, we propose a method to evaluate the pathogenesis of glaucoma that can be divided into three parts: the first is using random forest to segment the retinal nerve fiber layer; the second is using patch searching to segment disk and cup; and the third is calculating the relationship of retinal nerve fiber layer (RNFL) thickness and cup-to-disk ratio (CDR). Experimental results demonstrate that the correlations of the thickness of RNFL with vertical CDR, cup area and rim disk ratio are 0.64, 0.6, 0.54, respectively, which conforms that RNFL thickness and CDR are highly relative and complementary for the glaucoma diagnose. It plays an important role in the study on the development of glaucoma. : 017-01-19; : 017-03-14. 基金项目 : (616714); (3090140111004); (014-SWYY-04); () (MJUKF01706). 徐军 (1991 ),,, ; 陈强 (1979 ),,,,,, CCF, ; 牛四杰 (1984 ),,,, CCF,.
978 计算机辅助设计与图形学学报第 9 卷 Key words: optic coherence tomography; glaucoma; retinal nerve fiber layer; cup-to-disk ratio; random forest; support vector machine; correlation analysis 1 相关工作 表 1 视网膜各层结构信息,,.,, : 1) ; ) ; 3).,,.,., (optical coherence tomography, OCT),, OCT(SD- OCT),, [1]. 1 SD-OCT,, OCT,. OCT, 1. 1 SD-OCT OCT (optic nerve head, ONH),, RPE retinal pigment epithelium IS/OS inner and outer photoreceptor / segments ONL outer nuclear layer OPL outer plexiform layer INL inner nuclear layer IPL inner plexiform layer GCL ganglion cell layer NFL retinal nerve fiber layer ILM internal limiting membrane [-3].,,,,.,. 3 : 1) (intraocular pressure, IOP).,,. Stewart [4] (primary open-angleglaucoma, POAG), :,, POAG 5. Liza-Sharmini [5] 5 101 (chronic angle-closure glaucome, CACG) IOP, IOP CACG. ) SD-OCT. SD-OCT,,, RNFL,.., [6-7], [8-9], [10] [11] 4. 3)..
第 6 期, 等 : 979,,,. [1] SD- OCT RPE,.,.,. 3 : SD-OCT RNFL ; (support vector machine, SVM);. SD-OCT, SD-OCT,, Chen [13]. 本文算法,.,. 3,, RNFL..1 分割视网膜神经纤维层 4, OCT, B-scan, B-scan, A-line.. 3 5. 4 : Step1.., (region of interest, ROI) ILM RPE. Step.. Step1 ROI, A-line,. Step3. Step,. Step4. Step3,..1.1 OCT,., OCT,, [14]., 4 SD-OCT 5
980 计算机辅助设计与图形学学报第 9 卷 BM3D [15]. BM3D,. SD-OCT ILM RPE, ROI ILM RPE. ILM RPE,..1..,,,.,,, 6.. A-line,.,, 4.. 视觉神经头视盘视杯分割,.. Cheng [16],. SVM RPE. Hu [17] RPE,. 7 NCO OCT., NCO RPE, ; RPE 150 μm [18],, ILM. 6 7 OCT
第 6 期, 等 : 981 : Step1. RPE, OCT RPE RPE,,. Step. SVM RPE,., RPE ILM. 8. 8..1, OCT RPE RPE, RPE.1.1. RPE, RPE, RPE. RPE. RPE,,... SVM RPE RPE,. RPE. RPE, SVM. SVM libsvm,,,., RBF,,. 0 1 800 SVM, RPE. 4 : 1,, 3 RPE, 4.,,,, 81 81. LBP HOG SVM,..3 评估 RNFL 层厚度与杯盘比等相关参数的相关性 类 SD-OCT. 1 43, 1,, 18 104 51 B-scan ; 39, 19, 0, 00 104 00 B-scan. OCT Cirrus SD-OCT. :, ; RNFL. OCT RNFL,., RNFL, RNFL., 1, 3, 5, 10,,.,,,,,,. S disk, S cup, D disk, D cup, : =S cup /S disk, =D cup /D disk, S r =S disk S cup, = S r /S d. r r xy, N ( Xi X)( Yi Y) i 1 N N ( Xi X) ( Yi Y) i 1 i 1 n n n n xiyi xi yi i 1 i i n n n n i i i i i 1 i 1 i 1 i 1 n x x n y y 1,1. r 0, r 0, r ; r 0.5.,
98 计算机辅助设计与图形学学报第 9 卷 ~4 1,., 1~4 1, 3, 5, 10., RNFL,, 0.74, 0.61, 0.75 0.74; RNFL,, 0.8, 0.85, 0.79 0.8; RNFL,, 0.78, 0.75, 0.77 0.77. 表 左眼的相关性性分析 1 0.75 0.6 0.38 0.76 0.16 0.75 0.74 0.61 0.38 0.76 0.16 0.74 3 0.74 0.60 0.36 0.75 0.15 0.74 4 0.73 0.60 0.35 0.74 0.13 0.73 表 3 右眼的相关性性分析 1 0.81 0.85 0.37 0.79 0.19 0.81 0.81 0.84 0.36 0.78 0.17 0.81 3 0.8 0.85 0.36 0.79 0.13 0.8 4 0.83 0.86 0.35 0.80 0.17 0.83 表 4 第 1 类眼睛左眼 + 右眼的相关性分析 1 0.78 0.75 0.47 0.78 0.08 0.78 0.77 0.75 0.46 0.77 0.08 0.77 3 0.78 0.75 0.45 0.77 0.10 0.78 4 0.78 0.76 0.44 0.76 0.11 0.78 1, 5. 6 8. 表 5 第 类眼睛左眼 + 右眼的相关性分析 1 0.54 0.66 0.05 0.57 0.36 0.54 0.53 0.65 0.05 0.54 0.35 0.53 3 0.5 0.64 0.06 0.5 0.35 0.5 4 0.49 0.63 0.06 0.48 0.33 0.49 表 6 8 只眼睛相关性分析 1 0.54 0.64 0.14 0.6 0.17 0.54 0.53 0.63 0.14 0.6 0.16 0.53 3 0.53 0.63 0.13 0.61 0.16 0.53 4 0.53 0.64 0.11 0.61 0.19 0.53, RNFL,,. 3 总结与展望 NRFL, ; SVM, ; RNFL.,,,,. 参考文献 (References): [1] Wojtkowski M, Bajraszewski T, Targowski P, et al. Real-time in vivo imaging by high-speed spectral optical coherence tomography[j]. Optics letters, 003, 8(19): 1745-1747 [] Diekmann H, Fischer D. Glaucoma and optic nerve repair[j]. Cell and Tissue Research, 013, 353(): 37-337 [3] Son J L, Soto I, Oglesby E, et al. Glaucomatous optic nerve injury involves early astrocyte reactivity and late oligodendrocyte loss[j]. Glia, 010, 58(7): 780-789 [4] Stewart W C, Kolker A E, Sharpe E D, et al. Long-term progression at individual mean intraocular pressure levels in primary open-angle and exfoliative glaucoma[j]. European Journal of Ophthalmology, 008, 18(5): 765-770 [5] Liza-Sharmini A T, Yuen Shi Yin N, Shi-Huang Lee S, et al. Mean target intraocular pressure and progression rates in chronic angle-closure glaucoma[j]. Journal of Ocular Pharmacology and Therapeutics, 009, 5(1): 71-76 [6] Bagci A M, Shahidi M, Ansari R, et al. Thickness profiles of retinal layers by optical coherence tomography image segmentation [J]. American Journal of Ophthalmology, 008, 146(5): 679-687 [7] Shahidi M, Wang Z W, Zelkha R, et al. Quantitative thickness measurement of retinal layers imaged by optical coherence to-
第 6 期, 等 : 983 mography[j]. American Journal of Ophthalmology, 005, 139(6): 1056-1061 [8] Vermeer K A, van der Schoot J, Lemij H G, et al. Automated segmentation by pixel classification of retinal layers in ophthalmic OCT images[j]. Biomedical Optics Express, 011, (6): 1743-1756 [9] Lang A, Carass A, Hauser M, et al. Retinal layer segmentation of macular OCT images using boundary classification[j]. Biomedical Optics Express, 013, 4(7): 1133-115 [10] Yazdanpanah A, Hamarneh G, Smith B R, et al. Segmentation of intra-retinal layers from optical coherence tomography images using an active contour approach[j]. IEEE Transactions on Medical Imaging, 011, 30(): 484-496 [11] Niu Sijie, Chen Qiang, Lu Shengtao, et al. SD-OCT image layer segmentation method using multi-scale 3D graph search method[j]. Computer Science, 015, 4(9): 7-77(in Chinese) (,,,. SD- OCT [J]., 015, 4(9): 7-77) [1] 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, 015, 7(8): 1539-1547(in Chinese) (,,,. [J]., 015, 7(8): 1539-1547) [13] Chen X J, Niemeijer M, Zhang L, et al. Three-dimensional segmentation of fluid-associated abnormalities in retinal OCT: probability constrained graph-search-graph-cut[j]. IEEE Transactions on Medical Imaging, 01, 31(8): 151-1531 [14] Szkulmowski M, Gorczynska I, Szlag D, et al. Efficient reduction of speckle noise in optical coherence tomography[j]. Optics Express, 01, 0(): 1337-1359 [15] Chen Q, de Sisternes L, Leng T, et al. Application of improved homogeneity similarity-based denoising in optical coherence tomography retinal images[j]. Journal of Digital Imaging, 015, 8(3): 346-361 [16] Cheng J, Yin F S, Wong D W K, et al. Sparse dissimilarity-constrained coding for glaucoma screening[j]. IEEE Transactions on Biomedical Engineering, 015, 6(5): 1395-1403 [17] Hu Z H, Abramoff M D, Kwon Y H, et al. Automated segmentation of neural canal opening and optic cup in 3D spectral optical coherence tomography volumes of the optic nerve head[j]. Investigative Ophthalmology & Visual Science, 010, 51: 5708-5717 [18] Medeiros F A, Zangwill L M, Bowd C, et al. Evaluation of retinal nerve fiber layer, optic nerve head, and macular thickness measurements for glaucoma detection using optical coherence tomography[j]. American Journal of Ophthalmology, 005, 139(1): 44-55