Determination of the Intracranial Volume: A Registration Approach

Abstract

An algorithm to segment the intracranial compartment from PD-weighted MR images of the human head is described. If only a T1-weighted dataset is available for a given subject, an artifical PD-weighted dataset is computed from a dual-weighted reference by non-linear registration of the T 1-weighted datasets, and the intracranial compartment is segmented from this artificial dataset. The performance of the algorithm is evaluated on the basis of 12 dual-weighted data-sets with an average volume difference of 2.05% and an average overlap (Dice index) of 0.968. © Springer-Verlag 2004.