PDF version of this entire document
A common approach to assessment of the results of NRR involves the generation of test images. Such images are created by taking the original images and then applying known deformations to them. The process of evaluation is based on comparison between the deformation fields recovered by NRR and those which have originally been applied [19,21]. This type of approach can be used to test NRR methods 'off-line'. It cannot, however, be used to evaluate the results when the method is applied to real data as part of a registration-based analysis. Moreover, such artificial deformations fail to resemble real-world situations where there is an innate anatomical variation, which deformation are unable to capture. For instance, there may not be a one-to-one pixel/voxel relationship if images were acquired from different subjects. This property cannot be emulated by any fundamental deformation field.