In this paper we present an effective algorithmfor automated extraction of the vascular tree in retinal images,including bifurcations, crossovers and end-points detection.Correct identification of these features in the ocular fundushelps the diagnosis of important systematic diseases, such asdiabetes and hypertension. The pre-processing consists inartefacts removal based on anisotropic diffusion filter. Then amatched filter is applied to enhance blood vessels. The filteruses a full adaptive kernel because each vessel has a properorientation and thickness. The kernel of the filter needs to berotated for all possible directions. As a consequence, a suitablekernel has been designed to match this requirement. The maximumfilter response is retained for each pixel and the contrastis increased again to make easier the next step. A thresholdoperator is applied to obtain a binary image of the vasculartree. Finally, a length filter produces a clean and completevascular tree structure by removing isolated pixels, using theconcept of connected pixels labelling. Once the binary image ofvascular tree is obtained, we detect vascular bifurcations,crossovers and end points using a cross correlation basedmethod. We measured the algorithm performance evaluatingthe area under the ROC curve computed comparing the numberof blood vessels recognized using our approach with thoselabelled manually in the dataset provided by the Drive database.This curve is used also for threshold tuning.
|Number of pages||5|
|Publication status||Published - 2008|
All Science Journal Classification (ASJC) codes
- Biomedical Engineering