基于混合模型的视网膜血管自动分割算法
Retinal vessel automatic segmentation algorithm based on hybrid model

目的:构建基于混合模型的视网膜血管自动分割算法。方法和结果:混合模型算法流程包括4个步骤。首先,提取眼底图像的绿色分量图像以减少噪声影响(预处理); 然后分别应用形态学模型和尺度空间模型对预处理后的眼底图像进行分割; 将两种模型的分割结果进行融合; 最后,利用区域生长法对融合结果进行迭代生长,得到视网膜血管的精分割结果。从眼底图像库DRIVE训练数据集与测试数据集中分别抽取20幅彩色眼底视网膜图像进行自动分割,分割的准确度、敏感度和特异度分别为0.943 1、0.657 7、0.987 1。结论:混合模型算法克服了单一分割模型的局限性,能够获得较好的视网膜血管网络图像。
Aim:To establish a retinal vessel automatic segmentation algorithm based on hybrid model.Methods and Results:Hybrid model algorithm included 4 steps.Firstly, the green component image of the fundus image was extracted to reduce the noises influence(pretreatment); secondly,the morphological model and scale space model were respectively established to segment the preprocessed fundus images, and then the segmentation results of the two models were fused.Finally,the region growing method was applied to grow the fusion results iteratively, and the accurate segmentation results of the retinal vessels were obtained.A total of 40 color fundus retina images extracted from training data set(n=20)and test data set(n=20)of DRIVE library,were segmented using hybrid model,and the accuracy,sensitivity and specificity were 0.943 1, 0.657 7 and 0.987 1,respectively.Conclusion:Hybrid model algorithm can overcome the limitations of single segmentation model, and obtain a better retinal vascular network