The Reproducibility of Cerebrovascular Reactivity Across MRI Scanners

Olivia Sobczyk; Ece Su Sayin; Kevin Sam; Julien Poublanc; James Duffin; Joseph A Fisher; David J Mikulis

doi:10.3389/fphys.2021.668662

The Reproducibility of Cerebrovascular Reactivity Across MRI Scanners

Front Physiol. 2021 May 6:12:668662. doi: 10.3389/fphys.2021.668662. eCollection 2021.

Authors

Olivia Sobczyk^{1

2}, Ece Su Sayin^{2

3}, Kevin Sam⁴, Julien Poublanc¹, James Duffin^{2

3}, Joseph A Fisher^{2

3

5}, David J Mikulis^{1

5}

Affiliations

¹ Joint Department of Medical Imaging and the Functional Neuroimaging Laboratory, University Health Network, Toronto, ON, Canada.
² Department of Anaesthesia and Pain Management, University Health Network, Toronto, ON, Canada.
³ Department of Physiology, University of Toronto, Toronto, ON, Canada.
⁴ Department of Radiology, Johns Hopkins University School of Medicine, Baltimore, MD, United States.
⁵ Institute of Medical Sciences, University of Toronto, Toronto, ON, Canada.

Abstract

Cerebrovascular reactivity (CVR) is defined as the ratio of the cerebral blood flow (CBF) response to an increase in a vasoactive stimulus. We used changes in blood oxygenation level-dependent (BOLD) MRI as surrogates for changes of CBF, and standardized quantitative changes in arterial partial pressure of carbon dioxide as the stimulus. Despite uniform stimulus and test conditions, differences in voxel-wise BOLD changes between testing sites may remain, attributable to physiologic and machine variability. We generated a reference atlas of normal CVR metrics (voxel-wise mean and SD) for each of two sites. We hypothesized that there would be no significant differences in CVR between the two atlases enabling each atlas to be used at any site. A total of 69 healthy subjects were tested to create site-specific atlases, with 20 of those individuals tested at both sites. 38 subjects were scanned at Site 1 (17F, 37.5 ± 16.8 y) and 51 subjects were tested at Site 2 (22F, 40.9 ± 17.4 y). MRI platforms were: Site 1, 3T Magnetom Skyra Siemens scanner with 20-channel head and neck coil; and Site 2, 3T HDx Signa GE scanner with 8-channel head coil. To construct the atlases, test results of individual subjects were co-registered into a standard space and voxel-wise mean and SD CVR metrics were calculated. Map comparisons of z scores found no significant differences between white matter or gray matter in the 20 subjects scanned at both sites when analyzed with either atlas. We conclude that individual CVR testing, and atlas generation are compatible across sites provided that standardized respiratory stimuli and BOLD MRI scan parameters are used. This enables the use of a single atlas to score the normality of CVR metrics across multiple sites.

Keywords: MRI; carbon dioxide; cerebral blood flow; cerebrovascular reactivity; reproducibility.