R2’ is the Best Transverse Relaxation Rate for Oxygenation Mapping: Experience in Moyamoya Disease with Acetazolamide Challenge
Wendy Ni1,2, Thomas Christen2, and Greg Zaharchuk2

1Department of Electrical Engineering, Stanford University, Stanford, CA, United States, 2Department of Radiology, Stanford University, Stanford, CA, United States


Transverse MR spin relaxation rates, R2*, R2 and R2’ have all been considered sensitive to brain tissue oxygenation. In this study, we focus on a cohort of pre-operative Moyamoya disease patients and simultaneously map all three rates in addition to cerebral blood flow, both before and after the injection of the vasodilatory drug, acetazolamide. We found our measurements to be consistent with physiology and previous studies, and to support the use of R2’ for oxygenation mapping instead of R2* and R2.


Moyamoya disease is a progressive steno-occlusive disease causing chronic under-perfusion of brain tissues1. Transverse MR spin relaxation rates, R2* (=1/T2*), R2 (=1/T2), and R2’ (=R2*-R2), have been proposed for brain oxygenation mapping2. In this study, we simultaneous mapped these rates and cerebral blood flow (CBF), before and after the injection of acetazolamide (ACZ), to study each parameter’s change and correlation with disease state.


25 pre-operative Moyamoya disease patients (ages 38±12 y, 20 F) were scanned with informed consent and IRB approval, at 3.0T (MR750W, GE Healthcare) with ACZ challenge as part of a clinical diagnostic protocol. Relaxometry was performed using 2D Gradient-Echo Sampling of Free Induction Decay and Echo (GESFIDE: TESE/TR 100/2000ms, 40 echoes, TE 5-130ms, resolution 1.9×1.9×1.5mm3 with 34mm-thick coverage)3. CBF was measured using whole-brain 3D multi-delay pseudo-continuous ASL (TE/TR 25.1/6518ms, label time 2000ms, 5 equally spaced PLDs 700-3000ms, resolution 3.4×3.4×4mm3)4. Anatomical and angiographic information were acquired with whole-brain T1-weighted 3D IR-FSPGR images and intracranial time of flight (TOF) MRA.

GESFIDE images were analyzed in native space, and other images were coregistered to them. R2*, R2, and R2’ maps were calculated using mono-exponential fitting3. Tissue masks were segmented from anatomical images. Each dataset was analyzed in twelve 3cm annular, mixed-cortical ROIs approximating major arterial territories, with 6 radial segments located at 2 levels (Fig. 1). Each ROI was defined as angiographically “normal” and “abnormal” using blinded inspection of TOF MRA by an experienced neuroradiologist. Statistical significance of ACZ-induced changes and difference between groups was assessed using the 2-tailed t-test and α=0.05.


Results of group-level analysis are shown in Fig. 2-3. In abnormal ROIs, post-ACZ CBF was significantly lower by 12±38%, while pre- and post-ACZ R2’ were 13±32% and 14±33% higher, respectively. Pre- and post-ACZ R2 were 2.1±6.8% and 1.6±6.5% lower in abnormal ROIs.

ACZ-induced vasodilation augments CBF, increases tissue oxygenation, and reduces the amount of paramagnetic deoxyhemoglobin per voxel. Indeed, we observed that R2’, R2* and R2 all decreased significantly in normal tissues post-ACZ, by 2.5±13.8%, 1.3±3.1% and 0.2±2.6% (all p<0.05), respectively. However, Moyamoya disease-affected regions experience either submaximal or paradoxical reduction (“steal”) of CBF, which resulted in smaller ΔR2’, ΔR2* and ΔR2 reductions.

Finally, linear regression of each relaxation rate versus CBF (Fig. 4) found R2’ to be significantly negatively correlated with CBF (R2=0.16, p<0.05). R2 was uncorrelated (R2<0.01, p>0.05), while R2* showed very weak correlation (R2=0.05, p<0.05).


We observed R2’ trends consistent with physiology and a quantitative BOLD (qBOLD) biophysical model5 of oxygenation, for both ACZ- and disease-induced differences. qBOLD models tissue R2’ as proportional to the product of deoxygenated blood volume (DBV) and (1-SO2), where SO2 is the tissue oxygen saturation. ACZ-induced CBF augmentation causes competing effects via the two terms, but our observed increase in R2’ indicated that SO2 effect dominated, as reported by previous studies6. Higher pre-ACZ R2’ in abnormal ROIs may indicate higher DBV and/or lower SO2. However, without measuring pre- and post-ACZ blood volume in this cohort, these competing effects cannot be uncoupled. However, our results show that R2’ is sensitive to oxygenation changes, unlike R2*, which did not reflect tissue status, consistent with a prior study 7.

Modeling R2 is more complicated. There are two sources of R2 change: blood compartment and tissue. Like R2’, blood R2 is also affected by competing effects of oxygenation8 and blood volume9. In addition, due to small blood volume in tissues, even large blood R2 changes can be hard to measure with high SNR. Finally, tissue R2 can be changed by pathology independently from perfusion and oxygenation changes, e.g. in edema. Therefore, R2 is suboptimal for measuring tissue oxygenation.

To untangle the competing factors influencing R2’ and R2, accurate measurement of DBV is required before and after the ACZ challenge, which is challenging as most methods measure total cerebral blood volume and require contrast, which would not permit subsequent R2’ measurements. Non-contrast methods such as Vascular-Space-Occupancy (VASO)10 should be explored for this application.


In this study, we simultaneously characterized transverse spin relaxation rates R2*, R2 and R2’, with perfusion, in an ACZ challenge in Moyamoya disease patients. We observed consistent trends with what we expect from physiological models and prior studies, and physiology. We conclude that R2’ has the most potential for imaging oxygenation changes in tissue, with superior performance compared to R2* and R2.


NIH R01NS066506, R01NS047607, R21NS087491, NCRR 5P41RR09784. GE Healthcare.


1. R Piao et al., Ann Nucl Med, 2004.

2. T Christen et al., Am J Neuroradiol, 2013.

3. W Ni et al., Mag Reson Med, 2015.

4. W Dai et al., Mag Reson Med, 2012.

5. DA Yablonskiy & EM Haacke, Mag Reson Med, 1994.

6. H Okazawa et al., J Cereb Blood Flow Metab, 2001.

7. T Christen et al., Radiol, 2012.

8. JJ Chen & GB Pike, Mag Reson Med, 2009.

9. CM Anderson et al., Magn Reson Mater Phys, Biol Med, 2005.

10. J Uh et al., Mag Reson Med, 2009.


Fig. 1: Example of TOF MRA and regions of interest (ROIs) in a representative subject. This patient had a) unilateral Moyamoya disease affecting the right ACA and MCA regions (red arrows), and b-c) corresponding angiographic classification of 12 mixed-cortical ROIs.

Fig. 2: Group-level summary for angiographically normal (N=154) and abnormal (N=146) ROIs. * significant difference between normal and abnormal groups. We found that R2’ was lower and R2 was higher in abnormal regions, while R2* did not show any differences.

Fig. 3: Group-level summary of ACZ-induced changes. * significant difference between normal and abnormal groups. # significant change from baseline . We found that relaxation changes did not differentiate between normal and abnormal regions, but that larger decreases (corresponding to better oxygenation) were present in normal regions, as expected physiologically.

Fig. 4: Linear regressions of relaxation rates vs. CBF, pre- and post-ACZ, for all ROIs. R2’ in normal and abnormal ROIs also produced significantly different (p=0.01) lines not shown here: ynorm=-0.01x+3.47 and yabnorm=-0.02x+4.20. We found these results to confirm R2’ as the most sensitive rate for perfusion and oxygenation changes.

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)