Construction of a Brain Template from MR Images Using State-of-the-Art Registration and Segmentation Techniques

We propose a procedure for generating a brain atlas with mean morphology and mean intensity by state-of-the-art non-rigid registration of a database of MR images of normal brains. The new constructed atlas is much sharper that currently available linear atlases, as the residual inter-subject shape variability after both linear and subsequent non-linear normalization is retained. As a consequence, the resulting atlas is suited as a mean shape template for brain morphometry approaches that are based on non-rigid atlas-to-subject image registration.