基于高斯过程和Radon变换的视网膜血管管径测量

视网膜血管管径的异常变化与糖尿病、高血压等心脑血管疾病发展进程息息相关，眼底图像中视网膜血管信息的提取是计算机辅助分析和诊断相关疾病的重要步骤。本研究提出一种视网膜血管管径测量方法。首先对眼底图像进行图像预处理，然后基于高斯过程和Radon变换准确跟踪血管中心线和方向，最后利用二维高斯过程回归技术测量血管管径。在DRIVE和STARE这两个眼底图像数据库中进行测试。结果表明不论是对于曲率较小的近似直线型血管段、曲率较大的弯曲型血管段，还是对于管径发生变化的血管段，本文方法都能较好地检测出血管管径宽度，且标准差低、运算速度快。

Measurement of retinal vessel diameter using Gaussian process and Radon transform

The abnormal changes of retinal vessel diameter are closely associated with the development of cardiovascular and cerebrovascular diseases such as diabetes and hypertension. Therefore, the extraction of retinal vessel information in fundus images is significant for computer-aided analysis and diagnosis of related diseases. Herein a novel method for the measurement of retinal vessel diameter is presented. After fundus image preprocessing, Gaussian process and Radon transform are used to track the center lines and directions of retinal vessel, and then the retinal vessel diameter is estimated using two-dimensional Gaussian process regression technology. The fundus image databases of DRIVE and STARE are used for testing. The results reveal that whether for approximately linear vessel segments with small curvature, curved vessel segments with large curvature, or vessel segments with changes in diameter, the proposed method can measure the retinal vessel diameter efficiently, with low standard deviation and high operation speed.