Using Iterative Wavelet Cluster Analysis to Produce Individualised Reference Regions for Characterisation of Drift in Pharmacological MRIJohn McGonigle, Andrea L. Malizia, Robin Holmes, Robin Tyacke, Majid Mirmehdi, Using Iterative Wavelet Cluster Analysis to Produce Individualised Reference Regions for Characterisation of Drift in Pharmacological MRI. NeuroImage, Vol 47, pp. S–81. June 2009. PDF, 183 Kbytes.
MRI scanner drift is an important source of noise in BOLD experiments, with a time course usually exhibiting a lower frequency than most experimental paradigms and thus being easily removable. However, in pharmacological MRI (phMRI) experiments with slow acting agents – such as molecules acting on some widely distributed intracellular or genomic mechanisms – the brain’s response often has a time course similar to that of the scanner drift. This complicates efforts to remove it since approaches which presume any low frequency signal is drift will fail. Here we describe how the selection of a small volume with fewer receptors, and a data-driven approach to discover appropriate voxels within this, can be used to generate a clean reference region where sub-regions which appear to exhibit artefactual signal such as the effects of partial voluming may be excluded.