Abstract #2979

Zhang B, Gullapalli R

We propose a novel elastic registration algorithm which is based on the minimization of a physically motivated energy function, namely, the strain energy. Gauss-Seidel method was adopted in its numerical implementation. This algorithm was validated on prostate phantom magnetic resonance images. The registration accuracy was 1.00.6 pixels. Further validation was also performed on 6 sets of images obtained on patients undergoing endorectal prostate MRI examination. The normalized cross-correlation coefficient, which was used as a similarity measure between the reference and source/registered images, was significantly improved from 0.620.12 to 0.980.08 by the registration. These results demonstrated the feasibility of the registration algorithm to prostate images.