PMC

Vascular reactivity maps measured using: (a) breath-holding percentage signal change (BH), (b) resting-state physiological fluctuation regression (rs-PFR), and (c) resting-state coefficient of variation (rs-CV) techniques in a representative subject. CSF is masked in these images.

Respiratory volume and respiratory variation (RV) time series shown for two representative subjects: (a) a subject with good compliance with the breath-holding task and (b) a subject with poor compliance with the breath-holding task. The corresponding vascular reactivity maps, thresholded at z > 2.3, for these subjects are shown in (c) and (d).

(a) Schematic of the breath-holding paradigm. Each block was cued using a video projector, and repeated four times, followed by 26 s of normal breathing for a total scan time of 210 s. (b) Simulated respiratory volume and respiratory variation (RV) time series representing the situation in which the subject ideally complies with the breath-holding task.

The correlation coefficient values between the simulated and measured RV time series for all subjects are shown in (a). 6 of the 28 subjects, marked with red circles, had correlation coefficients below 0.2, demonstrating poor compliance with the breath-holding task. VR maps estimated using BH (b), rs-PFR (c), and rs-CV (d) methods for one of these subjects are presented. Only subjects with correlation coefficient > 0.2, marked with blue circles were included in comparative analyses.

Processing steps for (a) breath-holding and (b) rs-fMRI data. Breath-holding data were analyzed employing a 16-s ramp regressor delayed by 8 s and convolved with the standard SPM double gamma variate hemodynamic response function. rs-fMRI data were analyzed using two approaches: (1) Physiological fluctuation regression using GLM analysis of the average CSF signal (rs-PFR) and (2) voxel-based coefficient of variation (CV) of the signal fluctuation (rs-CV).
CSF: cerebrospinal fluid; GLM: general linear model.

Voxel-wise correlation values (R²) across all subjects between vascular reactivity as calculated with BH versus and the two rs-fMRI methods: (a) rs-PFR and (b) rs-CV, both of which were strongly statistically significant (P < 0.001). Additionally, the correlation between rs-PFR and rs-CV (c) was also significant (P < 0.001). Correlation values were calculated separately over the whole GM and WM voxels for each subject. The red line and whiskers, respectively, represent first and third quartiles, median, and minimum/maximum of the results in each individual.
rs-PFR: resting-state physiological fluctuation regression; rs-CV: resting-state coefficient of variation; BH: breath-holding.

Vascular reactivity measurements in different regions of interests (ROIs). rs-PFR and rs-CV versus BH measurements in 16 pre-defined structural ROIs are shown in (a) and (b), respectively, along with a color-coded map of their locations. A comparison of rs-PFR and rs-CV is shown in (c). Similarly, rs-PFR and rs-CV versus BH measurements in the GM and WM masks are shown in (d) and (e), respectively. A comparison of rs-PFR and rs-CV in the GM and WM ROI’s is also shown in (f). Both rs-PFR and rs-CV demonstrated statistically significant correlations with BH-derived VR, but the correlations for the rs-CV method tended to be higher.
rs-PFR: resting-state physiological fluctuation regression; rs-CV: resting-state coefficient of variation; BH: breath-holding; GM: gray matter; WM: white matter.