A method for removing global effects in small-animal functional MRI

Abstract

Global effects in functional MRI are temporal modulations in signal intensity resulting from various scanner and subject phenomena. These effects contribute to the overall variance, reducing the effect size associated with an experimental paradigm. Statistical estimations that include an approximation for concurrent global effects will reduce the residual error within the model and so improve statistical power of the study. Conventionally, estimates of global effects are derived from mean intracerebral signal intensities, but these may be prone to contributions from localised experimentally evoked signal changes. In such cases, inaccurate estimates of global effects may result in erroneous inferences of neural modulations based on statistical artefact. A novel, computationally simple, method of estimating global effects is proposed using muscle tissue acquired within the same acquisition volume. Quantitative improvements in sensitivity are reported for a somatosensory stimulation paradigm using global muscle signal intensities as a covariate of no-interest. The method is independent of local neurogenic signal changes and, under particular experimental conditions, may be more representative of true global effects. The utility of this strategy to applications in small-animal functional MRI that evoke systemic physiological changes as a result of the experimental manipulation is critically discussed.