Xi Liang1, 2, Kotagiri Ramamohanarao1, Qing Yang3, Alexander pitman4, Marius Staring5
1University of Melbourne, Carlton, Victoria, Australia; 2National ICT Australia, Carlton, Victoria, Australia; 3Apollo Medical Image Technology Pty. Ltd.; 4St. Vincent's Hospital; 5Leiden University Medical Center
DCE-MRI is sensitive in the detection of tumors. However the motion in between the image acquisitions can complicate the analysis. Nonrigid registration is used to achieve alignment between images. It can be formulated as an optimization problem to minimize the image dissimilarity. However it not only reduces the occurred motion but also may change the volume of enhanced regions, e.g. cancer tissue. A spatial-variant rigidity regularization term can be used to preserve the rigidity of tissues. This study proposes a practical method to compute the coefficients of rigidity terms. The evaluation result shows it can replace the manual segmentation to compute the coefficients used to reduce undesirable deformations. We propose a framework to generate regularization coefficients for nonrigid registration in DCE-MRI, where tumor locations are to be transformed in a rigid fashion. The coefficients are obtained by applying a sigmoid function on subtraction images from a pre-registration. All parameters in the function are automatically determined using k-means clustering. In the evaluation test, the proposed method can replace the binary coefficients requiring manual tumor segmentation.