Thalamus normalisation improves detectability of hypoperfusion via arterial spin labelling in an Alzheimer's disease cohort

Abstract

Data normalisation is an important approach to reduce inter-subject variability in group studies, but care must be taken when choosing a normalisation strategy to not introduce further confounds or artefacts into the data. Normalisation of arterial spin labelling perfusion measurements remains challenging, especially in the context of Alzheimer's disease, where there may be global hypoperfusion present. We propose that using the thalamus as a reference region for normalisation could improve the detectability of hypoperfusion in Alzheimer's disease and alleviate the pseudo-hyperperfusion artefacts caused by the commonly-used strategy of normalisation using global mean perfusion. Evaluation on an Alzheimer's disease dataset found this strategy was able to reduce coefficient of variation in perfusion measurements by around 60% and yield increases in statistical power of comparisons against healthy controls.