Segmentace optického disku (OD) a extrakce geometrických a intenzitních parametrů pro analýzu onemocnění sítnice

Abstract

The thesis deals with the problem of optic disc segmentation in retinal images and the subsequent analysis of its parameters for the purpose of diagnosing eye diseases. The main objective of the thesis is to develop a comprehensive system for accurate optic disc segmentation and extraction of its key features. Within the scope of the work, a segmentation model based on deep learning methods was developed, which was trained and optimized with respect to the diversity of clinical data. The model was subsequently tested on three different datasets in order to verify its robustness and accuracy. The segmentation part is followed by an algorithm for calculating geometric and intensity-based parameters of the optic disc. The extracted features were further used to design a machine learning classification model intended for pathology analysis. The achieved results show that the proposed approach enables effective segmentation and provides a basis for diagnostics. Thus, the thesis contributes to the automation of retinal image data processing.