Some files in the VBM2 toolbox were previously distributed as VBM-tools (cg_create_template.m / cg_vbm_optimized.m) and were based on modifications of code snippets of John Ashburner to implement the so-called optimized VBM protocol first proposed by Good et al. (2001). These files were extended by a couple of files in the toolbox to provide segmentation and statistical analysis of cross-sectional data as well as longitudinal data. The latter allow to track intra-subject effects over time by acquiring more than one scan per subject.
The key change to the SPM segmentation algorithms and the VBM-tools is the application of a Hidden Markov Random Field (HMRF) model. This model provides spatial constraints based on neighbouring voxels of a 3x3x3 cube. The idea is to remove isolated voxels of one tissue class which are unlikely to be member of this tissue type. This procedure also closes holes in a cluster of connected voxels of one tissue type. In the resulting segmentation the noise level will be minimized.