改进的多任务学习方法的眼底视盘分割与定位

doi:10.3969/j.issn.0255-8297.2021.06.006

摘要/Abstract

摘要： 提出了一种改进的多任务学习方法,网络的主结构由特征提取网络和分别进行视盘分割与视盘定位的双路径网络组成,通过端到端的训练与测试可以实现眼底图像视盘自动分割与定位相结合的多任务目的。在特征提取网络的编码阶段利用密集连接提取眼底图像视盘的上下文特征。视盘分割任务是依靠解码阶段逐步恢复原来的图像分辨率并获取整个视盘轮廓,视盘中心定位任务由空洞空间金字塔模块和金字塔池化模块来进一步提取视盘抽象特征,得到精准的视盘中心坐标。对350幅眼底图像进行了视盘分割和中心定位,实验结果表明:该方法自动分割的视盘结果与手动标注视盘区域的Dice系数为0.965,自动定位的视盘中心坐标与手动标记的视盘中心的平均绝对距离为0.191 mm(34.7像素)。

关键词: 视盘分割, 视盘定位, 密集块, 空间金字塔池化, 空洞空间金字塔池化, 多任务学习

Abstract: In this paper, an improved multi-task learning method is proposed. The main structure of proposed network contains a feature extraction network and a dual path network for optic disc segmentation and location respectively. And through end-to-end training and testing, the multi-task purpose of automatic optic disc segmentation and location can be achieved. In the decoding phase of the feature extraction network, a dense layer is used to extract the context features of fundus images. The optic disc segmentation task relies on the decoding stage to gradually restore original image resolution and obtain the entire optic disc outline, whereas the optic disc localization task is to obtain accurate optic disc center coordinates by extracting context features of fundus images with atrous space pyramid module and pyramid pooling module. Optic disc segmentation and center localization of 350 fundus images are demonstrated. Experimental results show that the Dice coefficient between the automatically segmented optic disc results and manually marked optic disc areas is 0.965, and the average distance between the automatic localization and the manually marked optic disc center is 0.191 mm (34.7 pixels).

Key words: optic disc segmentation, optic disc localization, dense block, spatial pyramid pooling, atrous spatial pyramid pooling, multi-task learning

中图分类号:

TP391

李宁, 尚英强, 熊俊, 邰宝宇, 时晨杰. 改进的多任务学习方法的眼底视盘分割与定位[J]. 应用科学学报, 2021, 39(6): 952-960.

LI Ning, SHANG Yingqiang, XIONG Jun, TAI Baoyu, SHI Chenjie. Fundus Optic Disc Segmentation and Localization Based on Improved Multi-task Learning Method[J]. Journal of Applied Sciences, 2021, 39(6): 952-960.

参考文献

[1] Zheng Y, Stambolian D, O'brien J, et al. Optic disc and cup segmentation from color fundus photograph using graph cut with priors[C]//Medical Image Computing and Computer-Assisted Intervention- MICCAI 2013. Berlin Heidelberg:Springer, 2013:75-82.
[2] Kubicek J, Timkovic J, Penhaker M, et al. Extraction of optical disc geometrical parameters with using of active snake model with gradient directional information[C]//Asian Conference on Intelligent Information and Database Systems. Cham:Springer, 2017:445-454.
[3] 侯向丹, 赵一浩, 刘洪普, 等. 融合残差注意力机制的UNet视盘分割[J]. 中国图象图形学报, 2020, 25(9):1915-1929. Hou X D, Zhao Y H, Liu H P, et al. Optic disk segmentation by combining UNet and residual attention mechanism[J]. Journal of Image and Graphics, 2020, 25(9):1915-1929. (in Chinese)
[4] Cheng J, Liu J, Xu Y, et al. Super pixel classification based optic disc and optic cup segmentation for glaucoma screening[J]. IEEE Transactions on Medical Imaging, 2013, 32(6):1019-1032.
[5] Mudassar A A, Butt S. Application of principal component analysis in automatic localization of optic disc and fovea in retinal images[J]. Journal of Medical Engineering, 2013(8):1-12.
[6] Sekhar S, Al-Nuaumy W, Nandi A K. Automated localisation of retinal optic disk using Hough transform[C]//IEEE International Symposium on Biomedical Imaging:From Nano To Macro. IEEE, 2008:1577-1580.
[7] Wu J, Liu M J, Xiao Z T, et al. Optic disc localization combining improved graph based visual saliency model and distribution feature of retinal blood vessels[J]. Journal of Medical Imaging & Health Informatics, 2018, 8(1):127-133.
[8] Mira P, Jesse S J, Luo S H. Locating the optic disc in retinal images[C]//International Conference on Computer Graphics, Imaging and Visualisation, 2006:141-145.
[9] Wu X, Dai B, Wei B. Optic disc localization using directional models[J]. IEEE Transaction Image Process, 2016, 25(9):4433-4442.
[10] Dehghani A, Moghaddam H A, Moin M S. Optic disc localization in retinal images using histogram matching[J]. EURASIP Journal on Image & Video Processing, 2012(19):1-11.
[11] 汪梓艺, 苏育挺, 刘艳艳, 等. 一种改进DeeplabV3网络的烟雾分割算法[J]. 西安电子科技大学学报, 2019, 46(6):52-59. Wang Z Y, Su Y T, Liu Y Y, et al. Algorithm for segmentation of smoke using the improved DeeplabV3 net-work[J]. Journal of Xidian University, 2019, 46(6):52-59. (in Chinese)
[12] 张博, 张苗辉, 陈运忠, 等. 基于空间金字塔池化和深度卷积神经网络的作物害虫识别[J]. 农业工程学报, 2019, 35(19):209-215. Zhang B, Zhang M H, Chen Y Z, et al. Crop pest identification based on spatial pyramid pooling and deep convolution neural network[J]. Transactions of the Chinese Society of Agricultural Engineering, 2019, 35(19):209-215. (in Chinese)
[13] 李英杰, 张惊雷. 基于全卷积网络的图像语义分割算法[J]. 计算机应用与软件, 2020, 37(8):213-218, 273. Li Y J, Zhang J L. Image semantic segmentation algorithm based on fully convolution network[J]. Computer Applications and Software, 2020, 37(8):213-218, 273. (in Chinese)
[14] 齐永锋, 侯璐璐, 段友放. 基于DenseNet-BC网络的皮肤镜下皮肤损伤分割[J]. 计算机工程与科学, 2020, 42(6):1060-1067. Qi Y F, Hou L L, Duan Y F. Dermatological skin lesion segmentation based on DenseNet-BC network[J]. Computer Engineering & Science, 2020, 42(6):1060-1067. (in Chinese)
[15] 黄同愿, 杨雪姣, 向国徽, 等. 基于YOLOV3的改进模型在行人检测中的应用[J]. 重庆理工大学学报(自然科学), 2020, 34(8):155-164. Huang T Y, Yang X J, Xiang G H, et al. Application of improved model based on YOLOV3 in pedestrian detection[J]. Journal of Chongqing University of Technology (Natural Science), 2020, 34(8):155-164. (in Chinese)
[16] 朱威, 绳荣金, 汤如, 等. 基于动态图卷积和空间金字塔池化的点云深度学习网络[J]. 计算机科学, 2020, 47(7):192-198. Zhu W, Shen R J, Tang R, et al. Point cloud deep learning network based on dynamic graph convolution and spatial pyramid pooling[J]. Computer Science, 2020, 47(7):192-198. (in Chinese)
[17] Dice L R. Measures of the amount of ecologic association between species ecology[J]. Ecology, 1944, 26(3):297-302.