Robustly measuring vascular reactivity differences with breath-hold: Normalising stimulus-evoked and resting state BOLD fMRI data

Inter-subject differences in local cerebral blood flow (CBF) and cerebral blood volume (CBV) contribute to differences in BOLD signal reactivity and, therefore, unmodelled variance in group level fMRI analyses. A simple way of elevating blood CO2 concentrations to characterise subject differences in vascular reactivity is through breath-holds but two aspects of this measure are often neglected: (1) breath-holds are usually modelled as blocks even though CO2 accumulates over time and (2) increases in CO2 differ between subjects. This study demonstrates that the BOLD breath-hold response is best modelled by convolving the end-tidal CO2 trace with a standard haemodynamic response function and including its temporal derivative. Inclusion of the BOLD breath-hold response as a voxel-dependent covariate in a group level analysis increases the spatial extent of activation in stimulus evoked and resting state datasets. By expressing the BOLD breath-hold response as a percentage signal increase with respect to an absolute change in the partial pressure of CO2 (expressed in mmHg), the spatial extent of stimulus-evoked activation is further improved. This demonstrates that individual end-tidal CO2 increases to breath-hold should be accounted for to provide an accurate measure of vascular reactivity resulting in more statistically active voxels in group level analyses. ►BOLD breath-hold response is best modelled using HRF-convolved end-tidal CO2 traces. ►Correcting stimulus-evoked responses with breath-hold measures increases significance. ►Subject differences in CO2 increases during breath-hold should be taken into account. ►Breath-hold normalisation can increase the extent of resting state correlations.