QUANTITATIVE EVALUATION OF DEFORMABLE IMAGE REGISTRATION

In this paper, a concept based on unbalanced energy is proposed to evaluate deformable registration algorithms. The new concept is demonstrated through the comparison of the ITK "demons" algorithm and thin-plate spline interpolation, performed on a prostate patient\´s CT images. The resultant displacement vector fields are substituted into a finite element based elastic framework to derive the unbalanced energy at each element. The detection of unbalanced energy coupled with a conditioned finite element model provides a quantitative assessment of each registration instance. The requirement of boundary constraints in a traditional mechanical model has been removed in the calculation of unbalanced energy. This method is automatic and suitable to evaluate deformable registration in a clinical setting.