Modelování sítnicového cévního systému s cílem jeho obrazové extrakce a standardizace cévní tortuozity

Abstract

This diploma thesis deals with complex analysis and modeling of the retinal vascular system. The aim of the thesis is the design and implementation of a fully automated segmentation model, allowing extraction of the blood system area within the binary model, where other retinal structures are suppressed. An important feature of the model is sensitivity and robustness to declare the effiency of segmentation in a environment with worse image parameters. For this reason, the designed model is also tested for data where the vascular system is visualized under low contrast. Part of the analysis is the comparitive testing of the designed model against selected segmentation methods based on objective criteria. Designed segmentation model allows the extraction of symptoms that evaluate the global degree of curvature of the vascular system. This procedure is key for the classification of pathological tortuosity, which i salso the subject of this work. The designed model is equipped with a registration procedure that allows to compensate geometric differences during repeated retinal scanning. The designed model was tested and verified on dataset from system RetCam 3, containing 44 images.