基于Gabor滤波器的糖尿病视网膜新生血管检测

目的:采用Gabor滤波器实现眼底图像中新生血管检测，帮助医生准确确定糖尿病视网膜病变的分期。 方法:对眼底图像进行预处理，并使用不同尺度参数和方向参数Gabor滤波器作用于预处理图像，并在尺度参数确定的情况下取各方向输出结果的最大值作为最后Gabor滤波器的输出。 结果:对比分析不同尺度参数的Gabor滤波器的结果，发现小尺度参数的Gabor滤波器在新生血管部分具有较强的输出。 结论:本研究提出的Gabor滤波器可以很好地区分眼底图像中正常血管与新生血管结构。

Gabor filter-based detection of neovascularization due to diabetic retinopathy

Abstract: Objective To achieve neovascularization detection in fundus images using Gabor filter for assisting doctors to determine the stages of diabetic retinopathy accurately. Methods Fundus images were preprocessed, and then the Gabor filters with different scale and directional parameters were applied to process the preprocessed images. The maximum value of the output in each direction was taken as the final output of Gabor filter on condition that the scale parameters were determined. Results The comparison of results obtained by Gabor filters with different scale parameters showed that Gabor filter with small scale parameters had strong outputs in neovascularization. Conclusion The proposed Gabor filter can accurately distinguish normal blood vessels and neovascularization in fundus images.