SW pro automatické modelování a evaluaci retinálních lézí

Abstract

One of the most serious eye diseases is retinopathy of prematurity, which affects premature babies. This disease is the most common cause of childhood blindness. For this reason, it is very important to monitor these neonatal patients, develop tools and find ways to detect the precursors and, in particular, the disease thresholds as effectively as possible. The practical part of the thesis focuses on the development of an algorithm that can automatically identify the morphological structure of retinal lesions and optical discs and enable their objective evaluation within one application. For evaluation and objective assessment of lesion size, a reference point was selected in the background image, which did not show significant changes in geometric properties throughout the patient's observation and against which individual retinal lesions were evaluated. Segmentation is based on the method of active contours, which can be imagined as deformable curves, which, by virtue of the action of energies, gradually change their shape until they reach the boundary of our segmented object. The resulting algorithm is applied to a user-friendly graphical-user interface that allows the results of geometric parameters to be further exported and archived for evaluation over time.