Fast feature based multi slice to volume registration using phase congruency

Slice to volume registration is very useful in many medical imaging applications, for example, fusing static high resolution three dimensional (3D) image volumes to dynamic two dimensional (2D) slice data for deriving motion information in 3D. Though information theoretic registration methods such as Mutual Information are usually robust, they are time intensive and typically require a high level of field-of-view correspondence between the source and target images. In single slice to volume registration scenarios, where such correspondence is limited, registration accuracy and robustness often deteriorate. In this paper, we present a novel registration method that maintains robustness and accuracy while significantly increasing registration speed. Our approach employs multiple slice (as opposed to single slice) to volume registration, which increases the amount of potential matching information while maintaining a small number of slices and hence facilitates the often necessary high speed dynamic image acquisition. Our proposed registration approach first extracts phase congruency information from the slices/volume using oriented 2D Gabor wavelets. Using local non maximum suppression, we then automatically obtain a robust and accurate set of feature points that are subsequently matched using an Iterative Closest Point (ICP) approach. Validation on BrainWeb simulated magnetic resonance imaging (MRI) data showed significant gains in speed (∼40-fold increase) when compared to conventional Mutual Information based volumetric registration while maintaining comparable robustness and accuracy levels.