PDF version of this entire document

Brain Dataset with Ground Truth

Our initial dataset consisted of $\mathcal{N}=36$ transaxial mid-brain 2D slices, extracted at equivalent levels from a set of T1-weighted 3D MR scans of normal subjects. The ground-truth data for this set consists of dense (pixel by pixel) binary tissue labels for the gray and white matter, the caudate nucleus tissue classes and CSF within the lateral ventricles. These labels were further divided into left and right. An example image and its labelling is shown in Figure 7.

Figure 6: Examples of the shuffle difference image: from one image to a second image (left), from the second image to the first (centre), and the symmetrical shuffle distance image (right)

The training set was non-rigidly registered using a Minimum Description Length (MDL) NRR algorithm [16]. This registration was used as the starting point for the evaluation.