Voxel-guided morphometry ("VGM") and application to stroke.

Monitoring of cerebral diseases associated with a change of morphology (e.g., stroke) requires unprecedented accuracy for quantification of its morphological progression for each voxel. The purpose of this paper is to provide a technique [voxel-guided morphometry (VGM)] to quantify macroscopic anatomical differences. VGM consists of four steps: 1) coarse linear alignment by the extended principle axes theory (ePAT) generalized to affine movements; 2) a cross-correlation-based technique using a matrix-norm for fine linear alignment; 3) the applied high-dimensional multiresolution full multigrid method determines the nonlinear deformations, thereby achieving a complete exploitation of information and effective processing. The method measures a gray-value-guided movement of each voxel from source to target. The resulting high-dimensional deformation field is further processed by 4) determination of volume alterations for each voxel. Furthermore, the effect of linear registration errors on final morphometric measurements is discussed and the conditions for a bijective correspondence of voxels assuming small alterations are derived. To illustrate the technique the changing morphology of different subjects suffering from cerebral infarction is presented by using commonly available T1-weighted magnetic resonance volumes. VGM visualizes that ischemic as well as remote regions are affected by stroke.