基于改进U型网络的眼底图像血管分割
|
摘要
眼底图像血管分割问题是眼科及其他相关疾病计算机辅助诊断的基础。通过分割和分析眼底图像中的血管结构,可以对糖尿病视网膜病变、高血压和动脉硬化等疾病进行早期诊断和监测。针对目前已有血管分割算法存在准确率不高和灵敏度较低的问题,基于深度学习基本理论,提出一种改进U型网络的眼底图像血管分割算法。首先,通过减少传统U型网络下采样和上采样操作次数,解决眼底图像数据较少的问题;其次,通过将传统卷积层串行连接方式改为残差映射相叠加的方式,提高特征的使用效率;最后,在卷积层之间加入批量归一化和PReLU激活函数对网络进行优化,使网络性能得到进一步的提升。在DRIVE和CHASE_DB1这两个公开的眼底数据库上进行实验,每个数据库随机抽取160 000个图像块送入改进的网络中进行训练和测试,可以得到该算法在两个数据库上的灵敏度、准确率和AUC(ROC曲线下的面积)值,相比已有算法的最好结果平均分别提高2.47%、0.21%和0.35%。所提出的算法可改善眼底图像细小血管分割准确率不高及灵敏度较低的问题,能够较好地分割出低对比度的微细血管。
The blood vessel segmentation of fundus images is the basis of computer-aided diagnosis of ophthalmology and other related diseases. Early diagnosis and monitoring of diseases such as diabetic retinopathy, hypertension and arteriosclerosis can be performed by segmenting the vascular structure in the fundus image. However, existing segmentation algorithms are challenged with low accuracy and low sensitivity. This paper proposed an improved U-Net fundus image segmentation algorithm based on the basic theory of deep learning. Firstly, the problem of less fundus data set was solved by reducing the number of pooling layers and upsampling layers of the traditional U-Net. Secondly, the use efficiency of the feature was improved by changing the traditional convolutional layer serial connection method to the residual mapping. Finally, the batch normalization and PReLU activation functions are added between the convolutional layers to optimize the network, which further improved the network performance. This paper conducted experiments on two public fundus databases, DRIVE and CHASE_DB1. The 160 000 image blocks were randomly extracted from each database and are sent into the improved network for training and testing. The sensitivity, accuracy and AUC(area under the ROC curve) of the algorithm were 2.47%, 0.21% and 0.35%, higher than those of the existing contrasted algorithms. The proposed algorithm improves the low accuracy and low sensitivity of small blood vessels segmentation in fundus images, and segmented small blood vessels better with low contrast.
The blood vessel segmentation of fundus images is the basis of computer-aided diagnosis of ophthalmology and other related diseases. Early diagnosis and monitoring of diseases such as diabetic retinopathy, hypertension and arteriosclerosis can be performed by segmenting the vascular structure in the fundus image. However, existing segmentation algorithms are challenged with low accuracy and low sensitivity. This paper proposed an improved U-Net fundus image segmentation algorithm based on the basic theory of deep learning. Firstly, the problem of less fundus data set was solved by reducing the number of pooling layers and upsampling layers of the traditional U-Net. Secondly, the use efficiency of the feature was improved by changing the traditional convolutional layer serial connection method to the residual mapping. Finally, the batch normalization and PReLU activation functions are added between the convolutional layers to optimize the network, which further improved the network performance. This paper conducted experiments on two public fundus databases, DRIVE and CHASE_DB1. The 160 000 image blocks were randomly extracted from each database and are sent into the improved network for training and testing. The sensitivity, accuracy and AUC(area under the ROC curve) of the algorithm were 2.47%, 0.21% and 0.35%, higher than those of the existing contrasted algorithms. The proposed algorithm improves the low accuracy and low sensitivity of small blood vessels segmentation in fundus images, and segmented small blood vessels better with low contrast.
引文
[1] 陆培, 王磊, 李志成, 等. 一种普适的基于多尺度滤波和统计学混合模型的血管分割方法[J]. 中国生物医学工程学报, 2016, 35(5): 519-525.
[2] 李居朋, 陈后金, 张新媛. 基于先验知识随机游走模型的视网膜血管分割方法[J]. 中国生物医学工程学报, 2009, 28(4):501-507.
[3] Ricci E, Perfetti R. Retinal blood vessel segmentation using line operators and support vector classification[J]. IEEE Transactions on Medical Imaging, 2007, 26(10): 1357-1365.
[4] Fraz MM, Remagnino P, Hoppe A, et al. An ensemble classification-based approach applied to retinal blood vessel segmentation[J]. IEEE Transactions on Biomedical Engineering, 2012, 59(9): 2538-2548.
[5] Fu H, Xu Y, Wong DWK, et al. Retinal vessel segmentation via deep learning network and fully-connected conditional random fields[C] //2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI). Prague: IEEE, 2016: 698-701.
[6] Orlando JI, Prokofyeva E, Blaschko MB. A discriminatively trained fully connected conditional random field model for blood vessel segmentation in fundus images[J]. IEEE Transactions on Biomedical Engineering. 2017, 64(1): 16-27.
[7] Liskowski P, Krawiec K. Segmenting retinal blood vessels with deep neural networks[J]. IEEE Transactions on Medical Imaging, 2016, 35(11): 2369-2380.
[8] 朱承璋, 邹北骥, 向遥, 等. 彩色眼底图像视网膜血管分割方法研究进展[J]. 计算机辅助设计与图形学学报, 2015,11: 2046-2057.
[9] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[J] //arXiv,2015,1505:04597
[10] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[11] Ngo L, Han JH. Multi-level deep neural network for efficient segmentation of blood vessels in fundus images[J]. Electronics Letters, 2017, 53(16):1096-1098.
[12] Gao X, Cai Y, Qiu C, et al. Retinal blood vessel segmentation based on the Gaussian matched filter and U-net[C]// 2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI). Shanghai:IEEE, 2017: 1-5.
[13] Roychowdhury S, Koozekanani DD, Parhi KK. Blood vessel segmentation of fundus images by major vessel extraction and subimage classification[J]. IEEE J Biomed Health Inform, 2017, 19(3):1118-1128.
[14] Marín D, Aquino A, Gegundez-Arias ME, et al. A new supervised method for blood vessel segmentation in retinal images by using gray-level and moment invariants-based features[J]. IEEE Transactions on Medical Imaging. 2011, 30(1): 146-158.
[15] Cheng E, Du L, Wu Y, et al. Discriminative vessel segmentation in retinal images by fusing context-aware hybrid features[J]. Machine Vision and Applications. 2014, 25(7): 1779-1792.
[16] Li Q, Feng B, Xie L, et al. A cross-modality learning approach for vessel segmentation in retinal images[J]. IEEE Transactions on Medical Imaging, 2016, 35(1): 109-118.
[17] Dasgupta A, Singh S. A fully convolutional neural network based structured prediction approach towards the retinal vessel segmentation[C] //2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017). Melbourne:IEEE, 2017: 248-251.
[18] Yan Z, Yang X, Cheng KTT. Joint segment-level and pixel-wise losses for deep learning based retinal vessel segmentation[J]. IEEE Transactions on Biomedical Engineering, 2018, 65:1912-1923.
[19] Azzopardi G, Strisciuglio N, Vento M, et al. Trainable COSFIRE filters for vessel delineation with application to retinal images[J]. Medical Image Analysis, 2015, 19(1): 46-57.
[20] Roychowdhury S, Koozekanani DD, Parhi KK. Iterative vessel segmentation of fundus images[J]. IEEE Transactions on Biomedical Engineering, 2015, 62(7): 1738-1749.
[21] Zhang J, Dashtbozorg B, Bekkers E, et al. Robust retinal vessel segmentation via locally adaptive derivative frames in orientation scores[J]. IEEE Transactions on Medical Imaging, 2016, 35(12): 2631-2644.
[2] 李居朋, 陈后金, 张新媛. 基于先验知识随机游走模型的视网膜血管分割方法[J]. 中国生物医学工程学报, 2009, 28(4):501-507.
[3] Ricci E, Perfetti R. Retinal blood vessel segmentation using line operators and support vector classification[J]. IEEE Transactions on Medical Imaging, 2007, 26(10): 1357-1365.
[4] Fraz MM, Remagnino P, Hoppe A, et al. An ensemble classification-based approach applied to retinal blood vessel segmentation[J]. IEEE Transactions on Biomedical Engineering, 2012, 59(9): 2538-2548.
[5] Fu H, Xu Y, Wong DWK, et al. Retinal vessel segmentation via deep learning network and fully-connected conditional random fields[C] //2016 IEEE 13th International Symposium on Biomedical Imaging (ISBI). Prague: IEEE, 2016: 698-701.
[6] Orlando JI, Prokofyeva E, Blaschko MB. A discriminatively trained fully connected conditional random field model for blood vessel segmentation in fundus images[J]. IEEE Transactions on Biomedical Engineering. 2017, 64(1): 16-27.
[7] Liskowski P, Krawiec K. Segmenting retinal blood vessels with deep neural networks[J]. IEEE Transactions on Medical Imaging, 2016, 35(11): 2369-2380.
[8] 朱承璋, 邹北骥, 向遥, 等. 彩色眼底图像视网膜血管分割方法研究进展[J]. 计算机辅助设计与图形学学报, 2015,11: 2046-2057.
[9] Ronneberger O, Fischer P, Brox T. U-net: Convolutional networks for biomedical image segmentation[J] //arXiv,2015,1505:04597
[10] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[11] Ngo L, Han JH. Multi-level deep neural network for efficient segmentation of blood vessels in fundus images[J]. Electronics Letters, 2017, 53(16):1096-1098.
[12] Gao X, Cai Y, Qiu C, et al. Retinal blood vessel segmentation based on the Gaussian matched filter and U-net[C]// 2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics (CISP-BMEI). Shanghai:IEEE, 2017: 1-5.
[13] Roychowdhury S, Koozekanani DD, Parhi KK. Blood vessel segmentation of fundus images by major vessel extraction and subimage classification[J]. IEEE J Biomed Health Inform, 2017, 19(3):1118-1128.
[14] Marín D, Aquino A, Gegundez-Arias ME, et al. A new supervised method for blood vessel segmentation in retinal images by using gray-level and moment invariants-based features[J]. IEEE Transactions on Medical Imaging. 2011, 30(1): 146-158.
[15] Cheng E, Du L, Wu Y, et al. Discriminative vessel segmentation in retinal images by fusing context-aware hybrid features[J]. Machine Vision and Applications. 2014, 25(7): 1779-1792.
[16] Li Q, Feng B, Xie L, et al. A cross-modality learning approach for vessel segmentation in retinal images[J]. IEEE Transactions on Medical Imaging, 2016, 35(1): 109-118.
[17] Dasgupta A, Singh S. A fully convolutional neural network based structured prediction approach towards the retinal vessel segmentation[C] //2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017). Melbourne:IEEE, 2017: 248-251.
[18] Yan Z, Yang X, Cheng KTT. Joint segment-level and pixel-wise losses for deep learning based retinal vessel segmentation[J]. IEEE Transactions on Biomedical Engineering, 2018, 65:1912-1923.
[19] Azzopardi G, Strisciuglio N, Vento M, et al. Trainable COSFIRE filters for vessel delineation with application to retinal images[J]. Medical Image Analysis, 2015, 19(1): 46-57.
[20] Roychowdhury S, Koozekanani DD, Parhi KK. Iterative vessel segmentation of fundus images[J]. IEEE Transactions on Biomedical Engineering, 2015, 62(7): 1738-1749.
[21] Zhang J, Dashtbozorg B, Bekkers E, et al. Robust retinal vessel segmentation via locally adaptive derivative frames in orientation scores[J]. IEEE Transactions on Medical Imaging, 2016, 35(12): 2631-2644.