Segmentační model pro automatickou detekci sleziny z CT obrazů a kvantifikaci radiofarmaka z PET obrazů

Abstract

The thesis is focused on the issue of a segmentation model for spleen detection on CT images and subsequent registration of spleen images on PET scans. This work is processed on real image data. To accurately detect the spleen from other tissues, image preprocessing was first performed and then spleen segmentation was compared to the gold standard. Segmentation was performed using edge-based and edgeless-based methods, with the edgeless-based method being much more efficient and accurate compared to the edge-based segmentation. Then, image registration was performed on PET scans, which allowed for accurate determination of radioisotope distribution in the spleen relative to the liver. The entire practical part of the work was created in the MATLAB computer environment and the subsequent results were realized in Microsoft Excel. After static evaluation, the results showed that the use of the given segmentation model and subsequent quantification of the radioisotope is beneficial for the diagnosis and treatment of subjects with spleen disease or pathology.