Analýza a modelování retinálního cévního systému z fluorescenční angiografie

Abstract

The main objective of this bachelor thesis is to test segmentation algorithms for the extraction of the vascular bed and the display of the time dynamics of filling of blood vessels with a contrast agent. A method based on the maximum principal curvature method and a method based on morphological operations were tested. Testing of algorithms was carried out using noise analysis, first segmenting images from fluorescence angiography, which later served as the gold standard, and then segmenting images with added Gaussian and Salt&Pepper noise. The noisy segmented images were compared to the gold standard using objectifying parameters. Graphs of the dependence of the selected parameter on noise level of image were compiled from the resulting objectification parameters, from which the quality of the algorithms was subsequently derived. A more suitable method for extracting the vascular bed from fluorescence angiography images turned out to be a method based on the maximum principal curvature method. To display the dynamics of filling blood vessels with contrast agent, image fusion was used for the registered images, which will highlight changes in time between individual images. For better visualization, gifs were subsequently created from these images, on which it is possible to observe the gradual filling of the vessels of the retina with a contrast agent.