Tensor-Based Morphometry

Tensor-based morphometry (TBM) is an image analysis technique that measures brain structural differences, cross sectional differences or changes over time in repeat scans, from the gradients of deformation fields that align one image to another. TBM may be applied to cross-sectional MRI data for local volumetric comparisons, based on nonlinearly registering individual brain scans to a common anatomical template (cross-sectional TBM). When TBM is applied in a longitudinal MRI study, a change map is computed by nonlinearly registering a follow-up scan to a baseline scan from the same individual (longitudinal TBM). TBM-derived measures of brain atrophy, reflecting the rate of tissue loss, can be used as an imaging surrogate biomarker to facilitate clinical trials. Care must be taken that the analysis methods are symmetric and free from multiple sources of bias [1]. Our method has been tuned to handle poorer quality scans robustly, i.e., it does not require the throw-out of scans, as a real clinical trial would not allow the selective exclusion of data.