Cerebrovascular reactivity measurements using 3T BOLD MRI and a fixed inhaled CO2 gas challenge: Repeatability and impact of processing strategy

Abstract

Introduction: Cerebrovascular reactivity (CVR) measurements using blood oxygen level dependent (BOLD) magnetic resonance imaging (MRI) are commonly used to assess the health of cerebral blood vessels, including in patients with cerebrovascular diseases; however, evidence and consensus regarding reliability and optimal processing are lacking. We aimed to assess the repeatability, accuracy and precision of voxel- and region-based CVR measurements at 3 T using a fixed inhaled (FI) CO2 stimulus in a healthy cohort.Methods: We simulated the effect of noise, delay constraints and voxel- versus region-based analysis on CVR parameters. Results were verified in 15 healthy volunteers (28.1±5.5 years, female: 53%) with a test-retest MRI experiment consisting of two CVR scans. CVR magnitude and delay in grey matter (GM) and white matter were computed for both analyses assuming a linear relationship between the BOLD signal and time-shifted end-tidal CO2 (EtCO2) profile.Results: Test-retest repeatability was high [mean (95% CI) inter-scan difference: −0.01 (−0.03, −0.00) %/mmHg for GM CVR magnitude; −0.3 (−1.2,0.6) s for GM CVR delay], but we detected a small systematic reduction in CVR magnitude at scan 2 versus scan 1, accompanied by a greater EtCO2 change [±1.0 (0.4,1.5) mmHg] and lower heart rate [−5.5 (−8.6,−2.4] bpm]. CVR magnitude estimates were higher for voxel- versus region-based analysis [difference in GM: ±0.02 (0.01,0.03) %/mmHg]. Findings were supported by simulation results, predicting a positive bias for voxel-based CVR estimates dependent on temporal contrast-to-noise ratio and delay fitting constraints and an underestimation for region-based CVR estimates.Discussion: BOLD CVR measurements using FI stimulus have good within-day repeatability in healthy volunteers. However, measurements may be influenced by physiological effects and the analysis protocol. Voxel-based analyses should be undertaken with care due to potential for systematic bias; region-based analyses are more reliable in such cases.

Cerebrovascular Reactivity in Patients With Small Vessel Disease: A Cross-Sectional Study

“…Full details of the MRI acquisition protocols including reproducibility can be found in previous works. 7 , 15 , 16 , 18 …”

“…We used linear regression to model the mean BOLD signal using a time-shifted end-tidal CO 2 profile and volume number (to account for linear signal drift) as independent variables. 7 , 18 We did not model voxel-wise BOLD signals as this lacks robustness against noise. 18 The optimal delay per subject and ROI was defined as the time-shift of the end-tidal CO 2 profile that gave the lowest sum of squared residuals.…”

“…Full details of the MRI acquisition protocols including reproducibility can be found in previous works. 7,15,16,18 During the 12-minute CVR scan, a physician or nurse was present, and we administered medical air and 6% CO 2 -enriched air (CO 2 :O 2 :N 2 , 6%:21%:73%) alternately for 2 and 3 minutes, respectively. 7 We monitored other physiological parameters: end-tidal CO 2 , end-tidal O 2 , oxygen saturation level, and heart and respiration rates.…”

Cerebrovascular reactivity to hypercapnia in patients with migraine: A dual‐echo arterial spin labeling MRI study

“…12,13 These MRI techniques have been tested to evaluate CVR in healthy participants, demonstrating good reproducibility. [14][15][16] A drawback of ASL is the low signal-to-noise ratio. BOLD MRI has the benefit of a higher signal-to-noise ratio but is on the other hand not a direct measure of CBF, making interpretation more difficult.…”

“…CVR can be measured as the response to a vasodilatory stimulus such as carbon dioxide (CO 2 ) by monitoring with MRI either the change in cerebral blood flow (CBF) using arterial spin labeling (ASL) or the change in blood oxygen level dependent (BOLD) signal 12,13 . These MRI techniques have been tested to evaluate CVR in healthy participants, demonstrating good reproducibility 14–16 . A drawback of ASL is the low signal‐to‐noise ratio.…”