3-D Registration Of Intermodality Medical Images
George I. Zubal, Hemant D. Tagare, Lei Zhang, James S. Duncan · 2005
We report a technique for registering functional and structural 3-D images of patient images (particularly applicable to the human brain). The registration is modeled as a solid rotation, translation and magnification of one image to fit the other. In increasing order of complexity, we developed a hierarchy of landmarks which consist of points, lines, curves, planes, surfaces and volumes. We allow for three different kinds of matches between landmarks: a complete match, a fragmented match, and a containment match. Our technique allows for explicit constraints on the possible rotation, translation and magnification obtained as a result of matching the landmarks. Finally, we show that the above can be cast as a linear programming problem and provide preliminary evidence that it leads to a fast and provably convergent algorithm for 3-D registration promising clinical registration accuracy on the order of one millimeter.