Purpose. To develop a technique to automate landmark selection for point-based interpolating transformations for nonlinear medical\r\nimage registration. Materials and Methods. Interpolating transformations were calculated from homologous point landmarks\r\non the source (image to be transformed) and target (reference image). Point landmarks are placed at regular intervals on contours\r\nof anatomical features, and their positions are optimized along the contour surface by a function composed of curvature similarity\r\nand displacements of the homologous landmarks. The method was evaluated in two cases (n = 5 each). In one, MRI was registered\r\nto histological sections; in the second, geometric distortions in EPI MRI were corrected.Normalizedmutual information and target\r\nregistration error were calculated to compare the registration accuracy of the automatically and manually generated landmarks.\r\nResults. Statistical analyses demonstrated significant improvement (P < 0.05) in registration accuracy by landmark optimization\r\nin most data sets and trends towards improvement (P < 0.1) in others as compared to manual landmark selection.
Loading....