A COMPARISON OF BRAIN IMAGING MODALITIES AND ANALYSIS APPROACHES FOR MEASURES OF CEREBROVASCULAR RESPONSIVENESS

Abstract

Cerebrovasculature responsiveness (CVR) has emerged as a key marker of brain health. However, recent findings report conflicting interpretations of CVR in healthy ageing, fitness and neurodegenerative conditions (e.g. dementia and stroke). This may be related to methodological differences (i.e. imaging modality, analysis approach and/or carbon dioxide (CO2) stimulus concentration). Within the same cohort, we examined the effect of: (1) imaging modality [functional magnetic resonance imaging (fMRI) vs. transcranial Doppler (TCD)]; 2) typical analysis approach (linear regression [fMRI] vs. 30-s mean steady state [TCD]), and 3) CO2stimulus concentrations, for determination of CVR. Thirty five healthy volunteers participated (20 young: 24 ± 7 years; 15 old; 66 ± 7 years). Participants completed two experimental sessions (TCD/fMRI) on separate days (randomised and counter-balanced). For both sessions, CVR data were obtained using two concentrations of CO2 via the same Douglas bag open circuit (4-min cycles of room air, 5% CO2, room air, 7% CO2). CVR measures were calculated for both fMRI and TCD data using typical analysis approaches (e.g. Driver et al., 2010 and Vernieri et al., 2009 respectively), and then applied to the other imaging modality for comparison. Differences between stimulus concentration (5% vs. 7% CO2) were also compared. There were no correlations (Pearson's r) between modalities for CVR (r<0.295, p>0.05; Fig 1). Within each modality, there were strong correlations between CVR values calculated via either analysis approach (r>0.737, p<0.001; Figs 2–3). Regardless of the modality, CVR derived via the TCD analysis approach (30-s steady state) was higher (Figs 4–5). Stimulus concentration altered CVR values for TCD (p>0.001, fig 5) but not fMRI modality (p=0.462, Fig 4), and the analysis approach used to derive CVR augmented this stimulus effect for the TCD modality (p=0.009, Fig 5). The analysis approach used to calculate CVR, as well as the CO2 concentration will have significant effects on reported values. Therefore, caution is needed when comparing studies that use different methodological approaches. Observed discrepancies between CVR measures across imaging modalities require further investigation to elucidate their cause. Calculating CVR values from fMRI and TCD data in 32 healthy individuals. Two different data acquisition and analysis methods (fMRI data with typical linear regression analysis vs. TCD data with typical 30-s steady state) with two different stimulus conditions (stimulus concentration: 5% vs. 7% CO2). Pearson's correlations revealed no correlations between the two modalities for either 5% CO2 (r = 0.295; p = 0.12) or 7% CO2 CVRs (r = 0.143; p = 0.46). Calculating CVR values from fMRI data in 32 healthy individuals. Comparing two different analysis approaches (fMRI: linear regression vs. TCD: 30-s steady state) applied to fMRI data acquired in two different experimental conditions (stimulus concentration: 5% vs. 7% CO2). Pearson's correlations revealed significant association of CVRs derived from each analysis approach for both 5% CO2 (r = 0.872; p < 0.001) and 7% CO2 (r = 0.956; p < 0.001). Calculating CVR values from TCD data in 35 healthy individuals. Comparing two different analysis approaches (fMRI: linear regression vs. TCD: 30-s steady state) applied to TCD data acquired in two different experimental conditions (stimulus concentration: 5% vs. 7% CO2). Pearson's correlations revealed significant associations of CVRs derived from each analysis approach for both 5% CO2 (r = 0.737; p < 0.001) and 7% CO2 CVRs (r = 0.789; p < 0.001). Calculating mean CVR values from fMRI data in 32 healthy individuals. Two different analysis approaches (fMRI: linear regression vs. TCD: 30-s steady state) in two different experimental conditions (stimulus concentration: 5% vs. 7% CO2). Two-way ANOVA revealed a significant main effect of analysis approach (p=0.003). The effect of stimulus concentration was not significant (p=0.462) and there was no interaction (p=0.638). Calculating mean CVR values from TCD data in 35 healthy individuals. Two different analysis approaches (fMRI: linear regression vs TCD: 30-s steady state) in two different experimental conditions (stimulus concentration: 5% vs 7% CO2). Two-way ANOVA revealed significant main effects of analytical approach (p > 0.001) and stimulus concentration (p > 0.001), and a significant interaction between analytical approach and stimulus concentration (p=0.009).