| Summary: | The distance transform is used for the preprocessing of image data for correlation methods of automatic registration of digital images. In our department the algorithm of automatic registration of images with the distance transform calculation by 4SSED+ method is used for the images of control and dynamic investigations, where there was a change in the position of the patient. In this thesis we analyze six best known distance transform algorithms on the simplified set of image data (4SSED, 4SSED+, SaitoXY, SaitoYX, Maurer, Eggers algorithms) and from those we choose the most suitable one for the purposes of automatic registration (4SSED+) and subsequently we use it in automatic registration of T1W images of a set of 76 MR examinations of head in patients with multiple sclerosis. This thesis proves that appropriately chosen and implemented algorithm of automatic registration with the use of distance transform can be as accurate as a manual method of registration and much faster. We performed the assessment of accuracy of 7 automatic registration with manual correction. From the nearly negligible correction values of subpixel shift and small rotation angles in the aforementioned set of 76 MR investigations and subsequently from their statistical characteristics the high accuracy of the used algorithm is apparent.
|
|---|
