Structural MRI and computational anatomy

Structural magnetic resonance imaging can yield highly detailed images of the human brain. In order to quantify the variability in shape and size across different brains, methods developed in the field of computational anatomy have proved exceptionally useful. For example, voxel-based morphometry is a popular method that involves segmenting magnetic resonance imaging scans into gray matter, white matter, and cerebrospinal fluid, and transforming individual brain shapes to a standard template space for comparative analysis. However, computational anatomy – when applied to brain data at scale – can be complex and computationally expensive. Furthermore, there are many possible pipelines that can be applied to structural brain data and for this reason it is important to follow best practices for reproducible neuroimaging analyses. This chapter demonstrates reproducible processing using the CAT12 (Computational Anatomy Toolbox) extension to SPM12 that focuses on voxel- and region-based morphometry. Through worked examples, we demonstrate three approaches to reproducible image analysis: ‘minimal’, ‘intermediate’ and a ‘comprehensive’ protocol using the FAIRly big workflow. The comprehensive approach automatically facilitates parallel execution of whole dataset processing using container technology and also produces re-executable run records of each processing step to enable fully automatic reproducibility.