Computer vision algorithms for magnetic resonance imaging
Ramin Zabih, Jun Hwan Kim · 2003
In this thesis we present two new algorithms for the post-processing of time series images obtained from contrast enhanced magnetic resonance angiography and a new algorithm for the correspondence problem for image pairs obtained from different modalities. The first algorithm is an automatic filtering algorithm using arterial characterization for selecting arterial phase images and mask images to generate an optimal summary arteriogram. Our algorithm use an evaluation function that qualitatively assesses how good a given subtraction image is. The evaluation function is fairly easy to implement and extremely fast to compute. We went through a paired double-blinded comparison to demonstrate that our algorithm is as effective as the manual process, but a lot faster than human counterpart. The second algorithm produces a segmentation of artery from vein/parenchyma in an input image sequence. Our algorithm uses a model-based time series analysis of individual pixels. We take an energy minimization approach to ensure spatial coherence. The energy is minimized in an expectation-maximization fashion that alternates between segmenting the image into a number of non-overlapping regions and finding the temporal profile parameters which describe the behavior of each region. Preliminary experiments on MR mammography and MR angiography studies show the algorithm's ability to find a reasonable segmentation. The third algorithm addresses visual correspondence problems without assuming that corresponding pixels have similar intensities. This situation is common with medical images, usually due to different imaging modalities. We use maximization of mutual information, which is a powerful technique for registering images that requires no a priori model of the relationship between intensities of corresponding pixels in different imaging modalities. Experiments demonstrate that our algorithm is effective for many situations, including real scenes that have had their intensities artificially altered.