基于形态学去噪的兔眼相干光断层成像中角膜形态自动识别

目的对兔眼相干光断层成像(optical coherence tornography,OCT)中的角膜边缘识别进行研究,以期从不具有较高清晰度的图像中自动获得角膜边缘以及相关形态学参数。方法首先利用Otsu算法对兔眼OCT图像进行二值化,并以半径为5像素的圆盘形结构元素进行形态学去噪运算消除内部伪边缘,通过Canny算子获得角膜边缘图像;然后使用筛选过错误边缘点后的其余边缘点进行边缘的二次曲线一般方程的曲线拟合,求出各点曲率半径;最后用克朗巴哈系数法与人工描绘的边缘获得的相应点的曲率半径进行一致性检验。结果基于形态学去噪的Canny算法获得了较为连续平滑的角膜内外边缘曲线,与手工标点拟合曲线求出的曲率半径值具有较高的一致性(Cronbach’sα=0.953,P0.05)。结论基于形态学去噪的Canny算法可以较为准确地对OCT图像中的兔眼角膜的形态进行识别。

Automatic recognition of corneal morphologyfrom optical coherence tomography of the rabbit eyes based on morphological denoising

Objective To automatically obtain corneal edges and associated morphological parameters from optical coherence tornography( OCT) of rabbit eyes. Methods Firstly,we used the Otsu algorithm for image binarization of rabbit eyes OCT images,and then, the circular structure element with radius of 5 pixels was used to perform morphological denoising in order to delete the internal false edges.Secondly,we obtained the corneal edge images by using Canny operator,then,error edge points were filtered.By fitting the extracted curves to the general equation of quadric curve, the curvature radius of each point on the cornea were obtained.Finally,the consistency checks between the curvature radius obtained by the algorithm in this paper and the curvature radius obtained by artificialwere carried out with Cronbach ' s Alpha. Results The Canny algorithm based on morphological denoising had extracted continuous and smooth corneal internal and external edge curves. The experimental results obtained higher consistency (Cronbach's α=0.953,P<0.05). Conclusions The Canny algorithm based on morphological denoising can more accurately recognize the form of rabbit corneas in OCT images.