3 - D Brain Image Registration Using Optimal Morphological Processing and Iterative Principal Axis Transform

S2 Sn S1 minimum distance minimum distance minimum distance fitness values 3−D positional variations ventricle 3−D shape 3−D structuring elements 3−D MST shape description 3−D MST shape description 3−D MST shape description Figure 1: Block-diagram representation of the 3-D GA optimization procedure. Figure 2: The slices of the optimal 3-D structuring element. able the best discrimination between ne 3-D positional variations of the ventricle shape. The conventional reproduction , crossover, and mutation procedures are extended to three dimensions. A variable crossover rate controlled by the population entropy is used. The block-diagram of the 3-D genetic algorithm optimization procedure is represented in Figure 1. The input shapes in Figure 1 are produced from a single shape by means of ne positional variations of the original 3-D shape. This is because the 3-D SE which best discriminates between ne positional variations of the single (original) 3-D shape has to be determined. The resulting SE is called the optimal SE. The slices of the optimal 3-D SE obtained by means of GA are shown in Figure 2. A set of IPAR and optimal MST matched images is shown in Figure 3. For a better graphical presentation to a user, a pseudocolored image is produced for each pair of registered MR and PET slices. The produced color images have the brightness determined by the intensity of MR gray images and the color determined by the brightness of PET gray images. Such a pseudocolored image is a convenient way of combining the anatomical information provided by MR imaging modality and metabolical information provided by PET imaging modality. A morphological method for improving the accuracy of the principal axis method for brain registration is presented. The morphological signature transform (MST) is used to describe ne positional variations of the brain ven-tricle. The method uses a genetic algorithm for selection of a near-optimal 3-D structuring element for MST-based registration. Experimental results demonstrate feasibility of the method. References [1] M.Bergstrom and et al, \Head xation device for reproducible position alignment in transmission ct and positron emmision tomography", Journal of Computer Assisted Tomography, vol. 5, pp. 136{141, 1981. [2] A.C.Evans and et al, \Mri-pet correlation in three dimensions using a volume-of-interest (voi) atlas", Journal of Cerebral Blood Flow and Metabolism, vol. 11, pp. A69{A78, 1991. [3] C.A.Pelizzari and et al, \Accurate three-dimensional registration of ct, pet, and/or mr images of the brain", registration: A new algorithm for retrospective correlation …