Mapping the end-tidal CO2 response function in the resting-state BOLD fMRI signal: Spatial specificity, test-retest reliability and effect of fMRI sampling rate

- We estimate the resting-state BOLD response to intrinsic PETCO2 fluctuations.
- We include CRV and RVT variations in the model to account for interaction between them and PETCO2.
- PETCO2 response has a peak around 7 seconds and its effect is more dominated in gray matter.
- fMRI sampling rate has minimal effect on the shape of the estimated responses.
- Test-retest reproducibility of the responses is higher for short-TR data.

The blood oxygenation level dependent (BOLD) signal measures brain function indirectly through physiological processes and hence is susceptible to global physiological changes. Specifically, fluctuations in end-tidal CO2 (PETCO2), in addition to cardiac rate variation (CRV), and respiratory volume per time (RVT) variations, have been known to confound the resting-state fMRI (rs-fMRI) signal. Previous studies addressed the resting-state fMRI response function to CRV and RVT, but no attempt has been made to directly estimate the voxel-wise response function to PETCO2. Moreover, the potential interactions among PETCO2, CRV, and RVT necessitate their simultaneous inclusion in a multi-regression model to estimate the PETCO2 response. In this study, we use such a model to estimate the voxel-wise PETCO2 response functions directly from rs-fMRI data of nine healthy subjects. We also characterized the effect of sampling rate (TR = 2 seconds vs. 323 ms) on the temporal and spatial variability of the PETCO2 response function in addition to that of CRV and RVT. In addition, we assess the test-retest reproducibility of the response functions to PETCO2, CRV and RVT. We found that despite overlaps across their spatial patterns, PETCO2 explains a unique portion of the rs-fMRI signal variance compared to RVT and CRV. We also found the shapes of the estimated responses are very similar between long- and short-TR data, although responses estimated from short-TR data have higher reproducibility.