Xiaofeng Liu1, Steven Roys2, Jerry L. Prince1,3, Rao Gullapalli2
1Computer Science, Johns Hopkins University, Baltimore, MD, USA; 2Radiology, University of Maryland Baltimore, Baltimore, MD, USA; 3Electrical and Computer Engineering, Johns Hopkins University, Baltimore, MD, USA
Magnetic resonance (MR) imaging and spectroscopy with an endorectal coil is a standard clinical procedure for prostate cancer diagnosis. However images and spectra are obtained when the prostate is in a distorted state due to the use of the endorectal coil. For optimal delivery of radiation dose to the cancerous tissue during therapy, these deformed images need to be transformed back to their non-deformed state. We propose an incompressible large deformation diffeomorphism method for landmark-based prostate image matching, which preserves the incompressibility, an important physical property of the prostate. Preliminary results on 2D images demonstrate the effectiveness of this method.