PDF version of this entire document

Data and NRR Experiments

In this particular set of experiments we deal with the IBIM dataset, which comprises 38 brain volumes (with labels), acquired from different subjects. The data is being registered onto a common frame of reference, using 4 different NRR algorithms, and its overlap [2] is being computed to verify that the algorithm leads to distinguishable results. A model can then be built from the registrations. NRR establishes a correspondence from which models of deformation can be built [3]. In this particular case, 4 appearance models of shape and intensity [1] are being built from the resultant correspondence. By measuring the quality of the models, the quality of the registration can be studied as well.