On the efficiency of the consistent retinal thick and thin blood vessels segmentation method

Author(s):  E.A. Fedotov, Belgorod State Technological University named after V.G. Shukhov, Belgorod, Russia

D.A. Chernomorets, Belgorod National Research University, Belgorod, Russia, daria013ch@yandex.ru

V.M. Mikhelev, candidate of Sciences, associate Professor, Belgorod National Research University, Belgorod, Russia

Issue:  Volume 45, №2

Rubric:  Computer simulation history

Annotation:  The analysis of state of the circulatory system’s vessels of the eye fundus is of great interest in the diag-nosis and treatment of various diseases. Segmentation of vessels and determination of their morphological features are the key stages of automated methods of diagnostic analysis of the vascular system, because the results of diagnosis depend on the accuracy of the selection and measurement of vessels elements. The authors analyzed the dependence of the results of the blood vessels segmentation on the images of the eye fundus from various partitions to pixel classes corresponding to thick and thin vessels that were obtained by k-means clustering. The analysis of the influence of the selected number of classes on the results of the segmentation of thick and thin vessels was carried out. The results of computational experiments using the eye fundus images from the openly accessible DRIVE database demonstrated the efficiency and effec-tiveness of the developed method of retinal blood vessels segmentation based on contrast limited adaptive histogram equalization, morphological filtering, k-means clustering and matched filtering.

Keywords:  eye fundus, blood vessels, segmentation, clustering, k-means method, Matlab, DRIVE database

Full text (PDF):  Download

Downloads count:  301