Moderní metody detekce a extrakce příznaků optického disku z retinálních obrazů

Abstract

The master thesis focused on testing the methods of automatic segmentation for detection of optical disc. Datasets Clarity RetCam 3, Envision RetCam 3, Phoenix ICON, Data Drive and Data Stare was used for the analysis. Altogether, 232 images of retina of adult patients and children diagnosed with retinopathy of prematurity (ROP) were analyzed. The first part was a manual processing of the images, where the golden standard came as a result. The core part was a proposal of algorithms for processing the data images. Firstly, an image was preprocessed using the method CLAHE and Bilateral filter. Active contours without the edges were used to make the segmentation algorithm. The method used freely deformed curves. The result received from this action was a binar imagge, wich was then compared to the golden standard. The differences between the methods were evaluated based on the evauation parameters. Therefore, the effecivity of methods of segmentation for different displaying modalities were evaluated. Finally, the graphical user interface (GUI) was made in the MATLAB program.