SW pro analýzu a modelování karcinomů prostaty a močového měchýře

Abstract

This diploma thesis deals with the design of an algorithm for the segmentation of prostate cancer and bladder from CT and MRI images. An accurate depiction and delineation of the prostate and bladder (critical organ) boundaries is a key radiotherapy prostate cancer problem. Physicians currently use manual segmentation. In clinical practice, this method is time-consuming and often affected by the skills and experience of physicians. The proposed method is based on an active contour segmentation, which works on the principle of a statistical distribution of pixel intensities in the image. The segmentation process of these organs will be started after delimiting the area of interest and the contour initializing the start point by the user, therefore, the proposed method is considered semi-automatic. The resulting contours of the individual organs are compared with manual segmentation created by the physician using the mean quadratic error, correlation, and DICE coefficient. In the next part of the testing, the area of automatically and manually created segments of the prostate cancer and bladder is calculated and compared. Thanks to these quantification comparisons and calculations the success of the used segmentation method is evaluated. The design and testing of this method were conducted in MATLAB. The practical output is a graphical user interface that shows the resulting segmentation. This segmentation can be exported, archived and further processed.