视网膜血管病变检测是辅助诊断高血压、糖尿病等诸多疾病的主要手段，定期的视网膜血管检测可以及时发现血管异常，帮助患者提前进行治疗，预防疾病的进一步发展和恶化。眼底视网膜血管结构复杂，轮廓模糊，交织分布，且不同图像之间差异性较大，血管分割和动/静脉识别难度较大。因此，采用图像处理的方法来自动的分析眼底视网膜图像，进行病情分析诊断具有十分重要的意义。

       针对眼底视网膜血管细小、结构复杂、轮廓模糊导致血管分割精度低的问题，提出一种多尺度框架下采用小波变换融合血管轮廓特征和细节特征信息的视网膜血管分割方法。通过预处理增强血管与背景的对比度，在多尺度框架下提取血管轮廓特征和细节特征，并分别进行图像后处理，采用小波变换融合两幅特征图像，计算各尺度对应像素的最大值，得到血管检测图像，采用Otsu法将血管二值化，实现视网膜血管的分割。

       针对视网膜动静脉小血管区分度较低导致整体动/静脉血管识别精度低的问题，提出一种基于提取血管子树分类的方法实现动/静脉血管的识别。利用视网膜血管的独特结构，在血管分割的基础上，提取血管中心线并检测出血管结构当中的关键点，提出一种投票检测算法分类出分叉和交叉，通过断开所有交叉点簇并采用8-邻域连通分量标记将所有血管划分为包含动脉或者静脉的多个子树，提取一组手工特征，利用K均值聚类方法对每个血管子树进行动/静脉标记，实现整体动/静脉血管的识别。

       基于DRIVE（Digital Retinal Images for Vessel Extraction）眼底图像数据库进行测试，血管分割的平均准确率、灵敏度和特异性分别为95.82%、70.86%和98.06%，动静脉识别的平均准确率为89.85%。测试结果表明，本文所提方法在血管分割和动静脉识别方面均取得较好的效果，对于辅助医生进行临床诊断具有一定的理论参考价值。

      Detection of retinal vascular disease is the main method to assist in the diagnosis of hypertension, diabetes and many other diseases. Regular retinal vascular detection can detect vascular abnormalities in time, help patients to treat them in advance, and prevent further development and deterioration of the disease. The retinal blood vessel structure of the fundus is complicated, the outline is blurred, and the distribution is intertwined, and the difference between different images is large, and the blood vessel segmentation and the arteriovenous recognition are difficult. Therefore, the use of image processing methods to automatically analyze retinal images for disease analysis and diagnosis is of great significance.

      Aiming at the problem of low blood vessel segmentation accuracy caused by the small, complex structure and blurred contour of the fundus retinal blood vessels, a retinal blood vessel segmentation method using wavelet transform to fuse the contour feature and detailed feature information of the blood vessel under a multi-scale framework is proposed. The contrast between the blood vessel and the background is enhanced by preprocessing, the contour feature and detail feature of the blood vessel are extracted in a multi-scale framework, and the image post-processing is performed respectively. The wavelet transform is used to fuse the two feature images, and the maximum value of the corresponding pixels of each scale is calculated to obtain the blood vessel. The blood vessel detection image is obtained, and the blood vessel is binarized by the Otsu method to realize the segmentation of the retinal blood vessel.

      Aiming at the problem that the low discrimination of retinal arterial/veinous small blood vessels leads to the low accuracy of the overall arterial/veinous blood vessel identification, a method based on the extraction of blood vessel subtree classification is proposed to realize the arterial/veinous vessel identification. Using the unique structure of the retinal blood vessels, the key points in the blood vessel structure are detected by extracting the centerline of the blood vessel, and a voting detection algorithm is proposed to classify the bifurcation and intersection clusters. All the blood vessels are marked by disconnecting all the intersection clusters and using 8-neighbor connected components divide into multiple sub-trees containing arteries or veins, extract a set of manual features and use K-means clustering method to label each vessel sub-tree with arterial/vein, so as to realize the recognition of the overall arterial/venous vessels.

      Based on the public DRIVE (Digital Retinal Images for Vessel Extraction) fundus image database, the research method was tested. The average accuracy, sensitivity and specificity of blood vessel segmentation were 95.82%, 70.86% and 98.06%, respectively. The average accuracy of arterial/veinous recognition is 89.85%. The test results show that the method proposed in this thesis has achieved good results in both blood vessel segmentation and arterial/veinous recognition, and has certain theoretical reference value for assisting doctors in clinical diagnosis.

[1]Kipli K, Enamul Hoque M, Thai Lim L, et al. Retinal image blood vessel extraction and quantification with Euclidean distance transform approach[J]. IET Image Processing,2020,14(15):3718-3724.

[2]Zou B, Fu H, Chen Z, et al. Ground truth free retinal vessel segmentation by learning from simple pixels[J].IET Image Processing,2021,15(6):1210-1220.

[3]国家卫生和计划生育委员会.“十三五”全国眼健康规划(2016-2020年)[J].中华眼科杂志,2017,53(07):484-486.

[4]王宁利,胡爱莲,汤欣,等.糖尿病视网膜病变分级诊疗服务技术方案[J].中华全科医师杂志,2017,16(08):589-593.

[5]Robbins C B, Thompson A C, Bhullar P K, et al. Characterization of retinal microvascular and choroidal structural changes in Parkinson disease[J].JAMA ophthalmology, 2021,139(2):182-188.

[6]Padmasini N, Umamaheswari R. Automated detection of multiple structural changes of diabetic macular oedema in SDOCT retinal images through transfer learning in CNNs[J]. IET Image Processing,2020,14(16):4067-4075.

[7]Sharifizad M, Schmidl D, Werkmeister R M, et al. Retinal vessel diameters, flicker‐induced retinal vasodilation and retinal oxygen saturation in high‐and low‐risk pregnancy[J].Acta ophthalmologica,2020.

[8]Xing W, Liu Y, Deng N, et al. Automatic identification of cashmere and wool fibers based on the morphological features analysis[J].Micron,2020,128:102768.

[9]El-Hag N A, Sedik A, El-Shafai W, et al. Classification of retinal images based on convolutional neural network[J].Microscopy Research and Technique,2021,84(3): 394-414.

[10]Chaudhuri S, Chatterjee S, Katz N, et al. Detection of blood vessels in retinal images using two-dimensional matched filters[J].IEEE Transactions on medical imaging,1989, 8(3):263-269.

[11]Farokhian F, Yang C, Demirel H, et al. Automatic parameters selection of Gabor filters with the imperialism competitive algorithm with application to retinal vessel segmentation[J].Biocybernetics and Biomedical Engineering,2017,37(1):246-254.

[12]Rezaee K, Haddadnia J, Tashk A. Optimized clinical segmentation of retinal blood vessels by using combination of adaptive filtering, fuzzy entropy and skeletonization[J]. Applied Soft Computing,2017,52:937-951.

[13]Gou D, Wei Y, Fu H, et al. Retinal vessel extraction using dynamic multi-scale matched filtering and dynamic threshold processing based on histogram fitting[J]. Machine Vision and Applications,2018,29(4):655-666.

[14]Singh N P, Srivastava R. Extraction of retinal blood vessels by using an extended matched filter based on second derivative of gaussian[J].Proceedings of the National Academy of Sciences, India Section A:Physical Sciences,2019,89(2):269-277.

[15]Dharmawan D A, Ng B P, Borijindargoon N. Design of Optimal Adaptive Filters for Two-Dimensional Filamentary Structures Segmentation[J].IEEE Signal Processing Letters,2019,26(10):1511-1515.

[16]Wang H, Jiang Y, Jiang X, et al. Automatic vessel segmentation on fundus images using vessel filtering and fuzzy entropy[J].Soft Computing,2018,22(5):1501-1509.

[17]Wang W, Wang W, Hu Z. Retinal vessel segmentation approach based on corrected morphological transformation and fractal dimension[J].IET Image Processing,2019, 13(13):2538-2547.

[18]Upadhyay K, Agrawal M, Vashist P. Unsupervised multiscale retinal blood vessel segmentation using fundus images[J].IET Image Processing,2020,14(11):2616-2625.

[19]Tian Y, Liu Z, Zhao S. Vascular segmentation of neuroimages based on a prior shape and local statistics[J].Frontiers of Information Technology & Electronic Engineering,2019,20(8):1099-1108.

[20]Zhao H, Li H, Cheng L. Improving retinal vessel segmentation with joint local loss by matting[J].Pattern Recognition,2020,98:107068.

[21]Liu I, Sun Y. Recursive tracking of vascular networks in angiograms based on the detection-deletion scheme[J].IEEE Transactions on medical imaging,1993, 12(2): 334-341.

[22]陈建辉,赵蕾,李德玉,万涛.基于冠脉造影图像血管树分割的血管狭窄自动识别方法[J].中国生物医学工程学报,2019,38(03):266-272.

[23]Deng W, Luo K, Shi Q, et al. Automatic Segmentation and Diagnosis Based on Multi-Scale Two-Stage Region Growing and Skeleton Extraction for Vessel Stenosis in Coronary Angiography[J].Journal of Medical Imaging and Health Informatics,2020, 10(2):446-451.

[24]Wang C, Oda M, Hayashi Y, et al. Tensor-cut: A tensor-based graph-cut blood vessel segmentation method and its application to renal artery segmentation[J].Medical image analysis,2020,60:101623.

[25]Yang J, Lou C, Fu J, et al. Vessel segmentation using multiscale vessel enhancement and a region based level set model[J].Computerized Medical Imaging and Graphics,2020,85:101783.

[26]Mo J, Zhang L. Multi-level deep supervised networks for retinal vessel segmentation[J].International journal of computer assisted radiology and surgery,2017, 12(12):2181-2193.

[27]Yan Z, Yang X, Cheng K T. Joint segment-level and pixel-wise losses for deep learning based retinal vessel segmentation[J].IEEE Transactions on Biomedical Engineering,2018,65(9):1912-1923.

[28]高宏杰, 邱天爽, 丑远婷, 等.基于改进U型网络的眼底图像血管分割[J].中国生物医学工程学报,2019,38(1):1-8.

[29]张赛,李艳萍.基于改进HED网络的视网膜血管图像分割[J].光学学报,2020,40(06):76-85.

[30]Lv Y, Ma H, Li J, et al. Attention guided U-Net with atrous convolution for accurate retinal vessels segmentation[J].IEEE Access,2020,8:32826-32839.

[31]Huang L, Liu F. Retinal vessel segmentation using simple SPCNN model and line connector[J].The Visual Computer,2020:1-14.

[32]殷宁波,黄冕,刘利军,黄青松.MS-UNet++:基于改进UNet++的视网膜血管分割[J].光电子·激光,2021,32(01):35-41.

[33]Franklin S W, Rajan S E. Computerized screening of diabetic retinopathy employing blood vessel segmentation in retinal images[J].Biocybernetics and Biomedical Engineering,2014,34(2):117-124.

[34]Barkana B D, Saricicek I, Yildirim B. Performance analysis of descriptive statistical features in retinal vessel segmentation via fuzzy logic, ANN, SVM, and classifier fusion[J].Knowledge-Based Systems,2017,118:165-176.

[35]Hassanien A E, Emary E, Zawbaa H M. Retinal blood vessel localization approach based on bee colony swarm optimization, fuzzy c-means and pattern search[J].Journal of Visual Communication and Image Representation,2015,31:186-196.

[36]Neto L C, Ramalho G L B, Neto J F S R, et al. An unsupervised coarse-to-fine algorithm for blood vessel segmentation in fundus images[J].Expert Systems with Applications,2017,78:182-192.

[37]Mirsharif Q, Tajeripour F, Pourreza H. Automated characterization of blood vessels as arteries and veins in retinal images[J].Computerized Medical Imaging and Graphics, 2013,37(7-8):607-617.

[38]Dashtbozorg B, Mendonça A M, Campilho A. An automatic graph-based approach for artery/vein classification in retinal images[J].IEEE Transactions on Image Processing, 2014,23(3):1073-1083.

[39]Hu Q, Abràmoff M D, Garvin M K K. Automated construction of arterial and venous trees in retinal images[J].Journal of Medical Imaging,2015,2(4):044001.

[40]薛岚燕,曹新容,林嘉雯,等.动静脉血管自动分类方法及其管径测量[J].仪器仪表学报,2017,38(09):2307-2316.

[41]黄文博.彩色眼底视网膜图像中相关目标检测方法研究[D].吉林:吉林大学,2018.

[42]Srinidhi C L, Aparna P, Rajan J. Automated method for retinal artery/vein separation via graph search metaheuristic approach[J].IEEE Transactions on Image Processing,2019, 28(6):2705-2718.

[43]Zhao Y, Xie J, Zhang H, et al. Retinal vascular network topology reconstruction and artery/vein classification via dominant set clustering[J].IEEE transactions on medical imaging,2019,39(2):341-356.

[44]Yin X, Irshad S, Zhang Y. Classifiers fusion for improved vessel recognition with application in quantification of generalized arteriolar narrowing[J].Journal of Innovative Optical Health Sciences, 2020,13(01):1950021.

[45]Remeseiro B, Mendonça A M, Campilho A. Automatic classification of retinal blood vessels based on multilevel thresholding and graph propagation[J].The Visual Computer, 2020:1-15.

[46]孙久琦.视网膜动静脉血管的识别及血管维数研究[D].北京:北京工业大学,2014.

[47]Lee M, Lee J G, Kim N, et al. Hybrid airway segmentation using multi-scale tubular structure filters and texture analysis on 3D chest CT scans[J].Journal of Digital Imaging, 2019,32(5):779-792.

[48]Zhang X, Zhang W. Application of new multi-scale edge fusion algorithm in structural edge extraction of aluminum foam[J].IEEE Access,2020,8:15502-15517.

[49]Prakash O, Park C M, Khare A, et al. Multiscale fusion of multimodal medical images using lifting scheme based biorthogonal wavelet transform[J].Optik,2019, 182:995-1014.

[50]Xing W, Liu Y, Deng N, et al. Automatic identification of cashmere and wool fibers based on the morphological features analysis[J].Micron,2020,128:102768.

[51]Lee M, Lee J G, Kim N, et al. Hybrid airway segmentation using multi-scale tubular structure filters and texture analysis on 3D chest CT scans[J].Journal of Digital Imaging, 2019,32(5):779-792.

[52]Kulkarni S C, Rege P P. Pixel level fusion techniques for SAR and optical images: A review[J].Information Fusion,2020,59:13-29.

[53]Yang R, Du B, Duan P, et al. Electromagnetic induction heating and image fusion of silicon photovoltaic cell electrothermography and electroluminescence[J].IEEE Transactions on Industrial Informatics,2019,16(7):4413-4422.

[54]Roychowdhury S, Koozekanani D D, Parhi K K. Iterative vessel segmentation of fundus images[J].IEEE Transactions on Biomedical Engineering,2015,62(7):1738-1749.

[55]Gou D, Wei Y, Fu H, et al. Retinal vessel extraction using dynamic multi-scale matched filtering and dynamic threshold processing based on histogram fitting[J].Machine Vision and Applications,2018,29(4):655-666.

[56]Ghoshal R, Saha A, Das S. An improved vessel extraction scheme from retinal fundus images[J].Multimedia Tools and Applications,2019,78(18):25221-25239.

[57]Zou B J, Chen Y, Zhu C Z, et al. Supervised vessels classification based on feature selection[J].Journal of Computer Science and Technology,2017,32(6):1222-1230.

[58]Muramatsu C, Hatanaka Y, Iwase T, et al. Automated detection and classification of major retinal vessels for determination of diameter ratio of arteries and veins[C]//Medical Imaging 2010:Computer-Aided Diagnosis. International Society for Optics and Photonics,2010,7624:76240J.