Komparativní studie segmentačních algoritmů pro identifikaci a extrakci příznaků cévního řečiště z variabilních medicínských akvizičních systémů

Abstract

This thesis is focused on creating a simulation graphical user interface in which is possible to compare segmentation methods and an influence of noise on the quality of image segmentation. Display of noise aplication and segmentation with subsequent extraction of blood vessels and pathology. An~integral part is also the calculation of evaluation parameters as objective indicators of segmentation quality. These indicators can then be plotted. The graphical user interface is created in MATLAB. The results are processed numerically and graphically. Noises selected for image data degradation were Gaussian noise, Salt & Pepper noise, Speckle Noise and Rician noise. Used segmentation methods were the Otsu multilevel thresholding and the Fuzzy C-means method. As~evaluation parameters were chosen SSIM (structural similarity index), MSE (mean square error), correlation coefficient, Dice coefficient, difference of area of extracted regions, calcification score and~its difference. The output is a report file containing graphical outputs of the batch noise application, batch segmentation, graphs of evaluation criteria and a structural variable of calculated values.