Cerebrovascular reactivity mapping using breath-hold BOLD-fMRI: comparison of signal models when using voxelwise lag optimization

We investigate the best modeling approach for mapping cerebrovascular reactivity (CVR) using voxelwise lag optimization. We considered two types of regressors (Block, PetCO2), three convolution models (no convolution, single gamma, double gamma), and a variable haemodynamic delay. We found that the CVR values obtained when using the PetCO2 signal without convolution, or convolved with a single gamma, are better than those obtained with the canonical HRF (double gamma, time-to-peak = 6s), while convolution with a canonical HRF remains the best option when using a Block design.