Severe intracranial arterial stenosis (SIAS) is a major public health issue and accounts for approximately 20% of all strokes¹. High oxygen extraction fraction (OEF) is a risk factor for developing an ischemic stroke in patients with SIAS and is usually measured by ¹⁵O₂-PET ^2,3. To increase availability and patient comfort, a non-invasive MR method for estimation of cerebral oxygen metabolism would be highly appreciated. A fast and robust measure of vascular deoxygenation, namely the relative OEF (rOEF) has been recently developed⁴ and successfully applied in patients with glioma⁵ and acute stroke⁶. However, in patients with subtle impairments, motion and other instabilities might impair results⁴. Here, we present preliminary data from a clinical study in patients with SIAS compared to healthy controls. The major aim was to evaluate the reliability of rOEF as well as Dynamic Susceptibility Contrast (DSC) and pseudo-Continuous Arterial Spin Labeling (pCASL) based perfusion measures by analyzing their hemispheric symmetry in gray matter (GM) in order to assess their potential diagnostic capabilities. In the ongoing clinical study, 15 subjects (9 healthy, 6 patients with unilateral SIAS >75%, age 70.6y±5.2y, 11 males) underwent MRI on a clinical Philips 3T Ingenia MR-Scanner (Philips Healthcare, Hamburg, Germany), using a 16ch head/neck-coil for clinical and 32ch head-coil for pCASL imaging. The protocol comprised rOEF-mapping (voxel size 2x2x3mm³, 112x92 matrix, 30 slices) by separate acquisition of a multi-echo GRASE (8 echoes, TE₁=ΔTE=16ms, TR=8971ms, acq.time 2:24min) and a multi-GE sequence (12 echoes, TE₁=ΔTE=5ms, TR=1950ms, α=30°, rapid flyback, acq.time 6:08min) for T2 and T2* mapping⁶. DSC data were obtained during a bolus injection of 15ml Gd-DTPA using single-shot GE EPI (TR=1516ms, TE=30ms, α=60°, 80 dynamics) after an carotid artery angiography with 17ml Gd-DTPA. Non-invasive perfusion mapping was performed by a single-shot pCASL sequence with a 2D EPI-readout (voxel size 3x3.1x5mm³, matrix 64x62, 16 slices, TE/TR=11ms/4396ms, label duration τ=1800ms, post labeling delay PLD=2000ms, background-suppression, 30 dynamics, acq.time 5:02min; including proton-density-weighted M₀ for normalization). Data coregistration, evaluation and calculation of rOEF ($$$rOEF\propto\frac{R2'}{rCBV}$$$ with $$$R2'=\frac{1}{T2*}-\frac{1}{T2}$$$) following Hirsch et al.⁶ and Cerebral Blood Flow (CBF) following the perfusion study group of ISMRM⁷ $$CBF=\frac{6000\cdot\lambda\cdot\Delta M\cdot e^{\frac{PLD}{T_{1}(Blood)}}}{2\cdot\alpha\cdot T_{1}(Blood)\cdot M_{0}\cdot\left(1-e^{-\frac{\tau}{T_{1}(Blood)}}\right)}$$ with brain-blood coefficient λ=0.9ml/g, label-control-difference ∆M, T1 of arterial blood at 3T T₁(Blood)=1650ms and labeling efficiency α=0.85 were performed with SPM12⁸ and custom Matlab programs⁹. All images were normalized to standard MNI brain and patient data were flipped so that the affected hemisphere was on the left side in each case. For each hemisphere, 9 VOI’s were defined (Fig.1), masked by GM (p_GM≥0.75) and mean values of the evaluated parameters calculated. Corresponding VOI averages of both hemispheres were compared by means of 2-sample T-tests and Bland-Altman plots¹².

All modalities showed good image quality and appeared reasonably symmetric on visual inspection (Fig.2). pCASL and DSC-based CBF maps looked similar but are not absolutely congruent. A closer, more quantitative analysis by means of Bland-Altman plots revealed that only pCASL-based CBF maps showed a detectable asymmetry for patients with SIAS with slightly decreased CBF on the affected hemisphere but not for healthy controls (Fig.3). DSC-based CBF and rCBV tended to be slightly asymmetric for both groups, whereas rOEF and DSC-based MTT are symmetric for both groups (Table 1). However, it has to be considered that the variances for all modalities are relatively high compared to the differences.

In healthy controls, our results point to a general symmetry of perfusion and oxygenation related measures between hemispheres (Table 1). Regarding patients, pCASL-based CBF implies a reduced perfusion on the side of carotid artery stenosis as expected (Fig.3). In contrast, qualitative DSC-based CBF does not show differences between both groups, which can be explained by general methodological shortcomings of DSC to estimate CBF¹⁴. Therefore, pCASL-based CBF maps are valued more reliable compared to DSC. Missing rOEF asymmetry in patients fits nicely with recent results of Bouvier et al. who found a decreased Cerebral Metabolic Rate for Oxygen ($$$CMRO_2\propto CBF\cdot OEF$$$), but symmetric tissular oxygen saturation in patients with arterial occlusion¹⁵. They assumed that decreased CMRO₂ was mainly caused by reduced CBF while OEF was unaffected which is in accordance with our preliminary results. Nevertheless, the presented method has to be further characterized with increasing the sample size.

These preliminary results are promising and show the expected behavior for healthy controls. In addition, pCASL CBF seems capable to detect perfusion impairments for patients with SIAS which is in accordance with recent literature. However, further measurements in a larger group of patients are required to increase statistical power and thereby proof the first findings.