Functional networks link grey and white matter cerebrovascular reactivity in healthy elderly adults

2604 

Abstract Submission 

Nuwan Nanayakkara¹, Liesel-Ann Meusel¹, Nicole Anderson^1,2, J. Jean Chen^1,2

¹Baycrest Health Sciences, Toronto, Ontario, ²University of Toronto, Toronto, Ontario, Canada

Nuwan Nanayakkara  
Baycrest Health Sciences
Toronto, Ontario

Liesel-Ann Meusel  
Baycrest Health Sciences
Toronto, Ontario

Nicole Anderson  
Baycrest Health Sciences|University of Toronto
Toronto, Ontario|Toronto, Ontario, Canada

J. Jean Chen  
Baycrest Health Sciences|University of Toronto
Toronto, Ontario|Toronto, Ontario, Canada

The cerebrovasculature was previously shown to be regulated in a network-like manner that coincides with the anatomy of cortical functional networks (Bright et al., 2020), consistent with early works that support the development of vasculature according to function (Shellshear, 1924). Thus, cerebrovascular reactivity (CVR), a hemodynamic metric often used to assess vascular regulation (e.g., diameter, resistance), may vary between these "vascular networks". Functional networks are fundamentally sustained by subcortical white-matter (WM) networks, and indeed, WM vascular density was found to vary by tract (Bernier, Viessmann and Ohringer, 2020), but the relationship between associated grey matter (GM) and WM hemodynamics remains unclear. Based on the literature on fetal development (Smirnov, Destrieux and Maldonado, 2021), we expect the vasculature feeding neighboring GM and WM regions within the same network to be congruent. Moreover, it is expected that the WM would be associated with longer hemodynamic delays than GM (Van den Bergh, 1969). This was previously reported for WM cerebrovascular reactivity (CVR) (Thomas et al., 2014). In this work, we investigate the relationships between CVR in paired GM and WM networks.

This study consisted of 18 healthy volunteers aged 71.42 years (mean). Each subject followed a 30 s resting and 2 s exhale followed by a 15 s breath-hold (period = 47 s), repeated 6 times during an dual-echo pCASL fMRI acquisition (voxel size = 3.4x3.4x6.0 mm, TE1/TE2/TR = 9.1/25/4000 ms). Respiratory bellows were recorded using the scanner's built-in belt and used as CVR delay reference. The time series data were preprocessed using FSL and AFNI tools. AFNI's 3dretroicor was used to generate BOLD data corrected for noise associated with physiological motion (i.e., heartbeat, respiration). Neighbouring control and tag frames were averaged in a sliding-window manner to produce 78 frames of BOLD data. These were then high-pass filtered at a 0.01 Hz cut-off frequency. CVR amplitude and delay were etimated using our Fourier-based method (Nanayakkara et al., 2023b), then divided into corresponding white matter and grey matter networks based on the functionnectome (Nozais et al., 2023). To minimize partial-volume effects, dilated FreeSurfer GM parcellations were used to separate GM from WM. The correlation between GM and WM of the CVR amplitude and time delay within each network was analyzed using Pearson correlation.

Fig. 1 shows the differences in mean CVR amplitude and time delay between GM and WM across the 30 resting-state networks. The CVR amplitude in WM was statistically significantly (p < 0.0001) lower and the time delay in WM was statistically significantly longer (p < 0.0001) than in GM as detected by paired t-test. These results are consistent with previous research (Thomas et al., 2014). Means and standard deviations in network-wise GM and WM regions exhibited a statistically significant correlation (Fig. 2). The mean time delay showed the strongest correlation between GM and WM.

This study was motivated by previous work by Bright et al. that demonstrated network behaviour in the BOLD response to carbon dioxide (Bright et al., 2020). Our results suggest CVR amplitudes are correlated across WM and GM regions within the same networks, but CVR delays even more so. The group-wise variability in these quantities are also coherent between WM and GM in the same network. BOLD-based CVR is determined by both vascular resistances (Duffin et al., 2018) and blood oxygenation, and CVR delay reflect both blood transit and the hemodynamic response time. For blood vessels in a given WM-GM network to exhibit similar CVR amplitude and delays, they should presumably exhibit similarities in some or all of the above factors that govern CVR.

Anatomy and Functional Systems ²

BOLD fMRI

Cerebral Metabolism and Hemodynamics ¹

Neurophysiology of Imaging Signals

Cerebral Blood Flow

White Matter