Comparison of two nonlinear registration techniques to investigate brain atrophy patterns in normal aging

SummaryBackground and purposeAutomated anatomical labeling (AAL) provides an automatic brain region segmentation method to allow objective measurement of regional brain volume. Nonlinear registration plays a critical role in such automated region-based volumetry. The aim of this study was to compare age-related brain regional volume changes using two nonlinear registration methods in statistical parametric mapping (SPM).Material and methodsThe study included 176 right-handed healthy participants (age range: 18–94 years). A total of 90 brain regions for each subject were automatically extracted, based on the AAL atlas, and two nonlinear registration methods (Normalization and DARTEL Toolbox in SPM5) were applied. Three-way ANOVA was performed to estimate the effects of brain region, each registration method and each hemisphere on regional volumes. Age-related brain-volume changes were also investigated by linear regression analysis for each nonlinear registration method.ResultsSignificant differences were found in volume among different brain regions (P < 0.001) with the two nonlinear registration methods (P = 0.011). Volumes of the corresponding brain region were significantly different (P = 0.037) between two hemispheres, and age-related volume reductions were unevenly distributed across regions. The most dramatic decreases in volume were found in the bilateral insula, middle frontal regions and cingulum. Rankings of the decreased brain regional volumes differed between the two registration techniques and adjustment methods.ConclusionThe inferred age-related volume atrophy patterns based on the AAL atlas were largely dependent on the choice of registration methodology.