Introduction

Cerebrovascular disease is an important contributor to dementia. Oxygen extraction fraction (OEF), cerebral blood flow (CBF), oxygen metabolism (CMRO₂), and arterial transit time (ATT) were measured using MRI in volunteers with a range of vascular disease risk (characterised by QRisk2 score) and cognitive impairment (characterised by Montreal Cognitive Assessment score) to determine if a) vascular disease risk was associated with altered vascular imaging parameters, b) if OEF was increased in association with impaired blood supply and increased transit times, and c) if cognitive impairment was associated with compensatory changes in CMRO₂, ATT, OEF, or CBF. We used Arterial Spin Labelling (ASL) to estimate CBF and ATT, and Quantitative Susceptibility Mapping (QSM) to estimate OEF from six veins in elderly subjects. We collected [¹⁸F]flutemetamol PET to determine amyloid status of each participant and so account for the contribution of amyloid positivity to cognitive impairment.

Methods

Twenty-four volunteers (age 69.5 ± 5.3 years; 14 female) with a range of vascular risk factors (QRisk2 18.7 ± 10.8 %) and cognitive impairment (MoCA 26.7 ± 3.4) were recruited.
MRI was performed using a 3T scanner (Signa PET/MR, GE Healthcare, Milwaukee, WI) and 32-channel head coil (Nova Medical, Wilmington, MA). QSM used a sagittal flow compensated 3D gradient multi-echo sequence: 1 mm isotropic voxels, 8 TEs (3 to 16.5 ms), TR 22.7 ms, flip angle 15°, ASSET acceleration (factor = 2, phase encoding direction) and bipolar echo collection. Multi-delay ASL with pCASL labelling, 3D spiral FSE read-out, 6 PLDs (700 to 2458 ms) and variable label duration (573-2042 ms) was collected with voxel size 1.7x1.7x4 mm. A T1w MP-RAGE image was collected with 1 mm isotropic resolution. Arterial blood saturation was estimated by sampling end-tidal O₂ using a nasal cannula. A blood sample was collected to measure haematocrit (Hct).
Susceptibility maps were calculated from the real and imaginary 3D gradient-echo images as described in Mc Fadden et al¹. Six veins (SS - Straight Sinus, ICVs – Internal Cerebral Veins, SSS || - Superior Sagittal Sinus (parallel to B₀), SSS ꓕ - Superior Sagittal Sinus (perpendicular to B₀), BVs – Basal Veins, TS – Tentorial Sinuses) were delineated using MRIcron (v6.6, https://people.cas.sc.edu/rorden/mricron/index.html). The 90^th percentile susceptibility value was used to calculate the venous oxygen saturation and then OEF in each region.
CBF and ATT maps were calculated following recommendations of the ASL white paper² using a single blood compartment model with fixed parameters T₁_blood =1.65 s, labelling efficiency=0.85, blood-brain partition coefficient=0.9 and voxel-wise M₀ calibration, fitting for both CBF and ATT.
Median CBF and ATT values were extracted from the vascular territory of the selected veins through co-registration (using FSL FLIRT) with the T1w images which were segmented using Freesurfer v6.0.0³.
CMRO₂ was calculated using:

$$CMRO_{2} = CBF \cdot [O_{2}]_{a}\cdot OEF$$

where [O₂]_a is the concentration of oxygen within arterial blood, calculated from Hct. Statistical analyses were conducted using R 3.6.0 (R Core Team, 2019). Three modelling strategies were used; for aims a) and b), linear mixed models and type III ANOVAs were used for each imaging parameter with subject as a random effect. For c) we averaged the regional values of the imaging parameters to give a global estimate. A linear model was constructed with MoCA score as the dependent variable and analysed using a type II ANOVA.

Results

ATT rose significantly by 85 ms per 10 point increase in QRisk (Table 1) with a trend for CBF to decline by 1.6 ml/min/100 ml for the same increase in QRisk. There was a significant interaction between amyloid status and QRisk for OEF, with amyloid positivity associated with a stronger association between OEF and QRisk. There was a trend for OEF to be lower in those with amyloid positivity, and for the relationship between OEF and QRisk to vary across regions.
OEF rose significantly by 5% for a decrease in CBF of 21.6 ml/min/100 ml or a rise in ATT of 432.0 ms (Table 2), with a trend for that relationship to vary by region (p = 0.084).
OEF and ATT increased with cognitive impairment (a rise of 16.8% in OEF and 703.2 ms in ATT for a 5 point reduction in MoCA score). Amyloid status was significantly associated with MoCA except for the ATT model, where a trend was observed.

Discussion

The significant relationship between ATT and QRisk, and trend between CBF and QRisk, are possibly due to increased cerebrovascular disease causing blockages and narrowing of blood vessels with all regions affected, though dysfunction further upstream (e.g. reduced cardiac output) could be contributing. The alterations in these two parameters with OEF confirm that oxygen extraction varies with cerebrovascular dysfunction. The links between OEF, ATT, and cognition suggests that the ability to compensate for reduced oxygen delivery with elevated OEF is insufficient to maintain cognitive function. CMRO₂ however, did not decline with QRisk and was not directly associated with MoCA. This may be because we are measuring CMRO₂ at rest when the deficiencies in oxygen supply are not apparent and metabolic demand can be met.