Automated optic nerve head detection in fluorescein angiography fundus images

The identification of the optic nerve head (ONH) is necessary preprocessing step in retinal image analysis, for automated extraction of the anatomical components in retinal images. In this study, a new image processing method based on Radon transform (RT) and multi-overlapping windows was proposed for ONH detection in fluorescein angiography (FA) fundus images. At first, RT was applied to all fundus sub images to find candidates for the location of the ONH. Then, the accurate location was found using the minimum mean square error estimation. The results of our automated method for the ONH detection in the images showed sensitivity and specificity of 90.54%, 98.51% respectively for pixel based analysis, and according to manual ONH detection, our automated algorithm found 89 ONH out of 100 in true location for FA images. This study addresses a novel method in detection of retinal land marks. Sensitivity and specificity of this algorithm seems to be acceptable in comparison with other detection methods.