眼底图像中硬性渗出物检测算法

目的利用眼底图像中硬性渗出物(hard exudates,HE)的亮度与边缘特征,提出一种基于Canny边缘检测算法与形态学重构相结合的HE自动检测方法,以解决目前算法灵敏度低、检测结果中视盘和血管的干扰等问题,对糖尿病视网膜病变(diabetic retinopathy,DR)的自动筛查具有重要意义。方法检测算法包括4个步骤。步骤一,图像预处理,主要包括RGB通道选取、基于形态学的图像对比度增强。步骤二,视网膜图像关键结构的消除,利用基于Gabor滤波的血管分割方法,消除血管边缘对HE检测的影响。将本文视杯分割算法应用在眼底图像红色通道上实现视盘自动分割,消除视盘及其边缘对HE检测的影响。步骤三,利用改进的Canny边缘检测算法和形态学重构方法对HE进行提取。步骤四,基于形态学的图像后处理,消除眼底图像边缘部分假阳性区域。最后利用该算法测试公开数据库中的40幅图像(35幅HE病变图像,5幅正常图像)。结果该算法对基于病变的灵敏性(sensitivity,SE)和阳性预测值(positive predictive value,PPV)分别为93.18%和79.26%,基于图像的灵敏性、特异性(specificity,SP)和准确率(accuracy,ACC)分别为97.14%、80.00%和95.00%。结论与其他方法对比,基于Canny边缘检测算法与形态学重构相结合的HE自动检测算法具有较好的可行性。

Algorithm of hard exudates detection in fundus image

Objective By using the characteristics of brightness and edge in hard exudates ( HE) ,a new automatic HE detection method based on a Canny edge detection algorithm combined with morphological remodeling method is proposed. The purposes are to solve problems like low sensitivity and interference of optic disk and vessels in other algorithms. It has great significance to the automatic screening of DR. Methods This algorithm could be divided into four steps. The first step was the image preprocessing, mainly including the selection of RGB channels and image contrast enhancement based on morphology. The second step was the elimination of key structures in retinal image. To avoid interference to the HE detection, the vessel segmentation method based on Gabor filter was used to eliminate the influence of blood vessel edge,and the proposed optic cup segmentation method was used in red channel to eliminate the optic disk and its edge. The third step was to extract the HE by using the improved Canny edge detection operator combined with morphological reconstruction method. The forth step was image post-processing based on morphology,eliminating the false positive area in image edges. At last,we tested 40 images in the public database(35 images with HE lesions,5 normal images). Results Its lesion based sensitivity and positive predictive value were 93. 18％,79. 26％,respectively. Its image based on sensitivity,specificity and accuracy were 97. 14％,80. 00％ and 95. 00％,respectively. Conclusions Comparing the above evaluation indexes with other methods, and the results proved the feasibility of the algorithm based on a Canny edge detection algorithm combined with morphological remodeling method.