Automatic segmentation of brain structures is an important prerequisite for many applications in neuroscience. In this paper a fully automatic method for segmenting the caudate from 3D MRI brain images is presented. The method is based on multi-atlas registration which has shown to be a powerful concept for segmentation. The results show that the automatic segmentation is similar to segmentations by human observers for routine data and slightly worse but still acceptable for nonroutine data.
A multi-atlas approach to automatic segmentation of the caudate nucleus in MR brain images
E. van Rikxoort, Y. Arzhaeva and B. van Ginneken
3D Segmentation in the Clinic: A Grand Challenge 2007:29-36.