Automated registration of CT and MR brain images using 3-D edge detection

Presents work on automatic inter-modality image registration based on the extraction of the most well-known image feature-edges. While many of the current techniques use the corresponding surface feature (skin surface and brain surface in particular) for image registration, an edge feature can include ail of those surfaces as well as certain tissue structures. The edge feature may provide as well the simplicity and possibility for surface and contour information to be further constructed. In contrast to most of the modern surface registration algorithms which rely on either manual or automatic segmentation of the corresponding surface, the authors' 3-D edge registration algorithm requires no further segmentation for the registration of CT and MR brain images. The registration accuracy is evaluated through a fiducial "gold-standard" on 6 pairs of CT and MR (T1, T2, and PD) images. The preliminary results are comparable to those obtained by other registration techniques.