I BOLD CO₂ MR Measurement of CVR

Blood oxygen level–dependent (BOLD) MRI refers to a MRI acquisition that is sensitive to the concentration of deoxyhemoglobin in blood. Arterial blood is normally 95% to 100% saturated with oxygen, and venous blood is approximately 60% to 70% saturated. An increase in CBF during stable neuronal activity (i.e., a stable number of deoxyhemoglobin molecules produced per unit time) results in dilution or “wash-out” of deoxyhemoglobin in the cerebral microcirculation. Deoxyhemoglobin is a paramagnetic substance. Paramagnetic substances constrained by boundaries such as red cell membranes or blood vessel walls yield magnetic field inhomogeneities that act to reduce the MRI signal. Therefore, an increase in CBF in the setting of constant neural activity results in a decrease in deoxyhemoglobin concentration in the microcirculation, yielding an increase in MR signal. Furthermore, % BOLD MRI signal is linearly related to % CBF over the range of CBF changes that occur in a CVR study,¹⁶ enabling the use BOLD MRI to measure CVR.

Advantages of the BOLD MRI technique are the use of an MR pulse sequence that is standard on modern clinical MR systems, short duration of study (acquisition time less than 10 minutes, and real-time data analysis), no requirement for intravenous injection, no ionizing radiation, and the ability to acquire routine MRI and magnetic resonance angiography (MRA) of the brain in the same imaging session. This technique has particular potential when one uses CO₂ as the vasodilatory stimulus: unlike the sustained effect of acetazolamide, arterial pCO₂ is rapidly responsive to changes in the inhaled concentration of CO₂. The BOLD MRI pulse sequence, which is based on rapid echoplanar imaging, can acquire a complete data set covering the entire brain in 2 seconds. The sequence is repeated several hundred times over the course of several minutes in order to provide statistical power for the data analysis while correcting for imaging system instabilities.

The greatest concern with BOLD CVR has been that BOLD MRI signal depends on CBF, but also (and to an unknown extent) on other factors: cerebral blood volume (CBV), cerebral metabolic rate of oxygen consumption (CMRO₂), arterial partial pressure of oxygen (PaO₂), and hematocrit. However, we now have empirical evidence that these other factors are not major issues. Goode et al.¹⁷ and Kassner et al.¹⁸ have demonstrated between-session reproducibility of BOLD CO₂ CVR, and Shiino et al.¹⁹ and our group²⁰ have demonstrated the accuracy of BOLD CO₂ CVR in patients with severe cerebrovascular disease. Shiino et al. compared BOLD CO₂ CVR with SPECT acetazolamide CVR in 10 patients with severe carotid stenosis or occlusion. They found that the degree of impairment and distribution of impaired areas detected by BOLD MRI correlated with the results of SPECT. Our group studied 38 patients with steno-occlusive disease by both BOLD CO₂ CVR and arterial spin labeling (ASL) MRI. The latter technique does not have the CBV and CMRO₂ dependencies of BOLD MRI. Hemispheric CVR measured by BOLD MRI was significantly correlated with that measured by ASL MRI for both gray matter (r = 0.83, p < 0.0001) and white matter (r = 0.80, p < 0.0001). Diagnostic accuracy (area under receiver operating characteristic curve) for BOLD MRI discrimination between normal and abnormal hemispheric CVR was 0.90 (95% CI = 0.81 to 0.98; p < 0.001) for gray matter and 0.82 (95% CI = 0.70 to 0.94; P<0.001) for white matter. Regions of paradoxical CVR on BOLD MRI had a moderate predictive value (14 of 19 hemispheres) for spatially corresponding paradoxical CVR on ASL MRI. Complete absence of paradoxical CVR on BOLD MRI had a high predictive value (31 of 31 hemispheres) for corresponding nonparadoxical CVR on ASL MRI.

The BOLD CO₂ CVR examination

There are a variety of ways to perform a BOLD CO₂ CVR study. Our technique has evolved over time, and we will discuss our current technique in detail.

Patient Preparation: We have performed BOLD CO₂ CVR studies in over 400 patients with the most severe forms of cerebrovascular disease. pCO₂ levels are used that are similar to the range of pCO₂ that occurs during daily living. Fewer than 5% of patients have not been able to complete the examination. Outside the MR scan room, we place a soft plastic facemask on the patient, and use plastic tape (3M Tegaderm™) to ensure an air-tight seal to the face. The patient is then taken into the MR scan room, and positioned supine on the scanner table with head inside the head coil.

Vasodilatory Stimulus: Patients’ expiratory gas concentrations are controlled by providing a series of gas flows with variable CO₂, O₂, and N₂ concentrations to a sequential re-breathing manifold²¹ (Thornhill Research Inc., Toronto, Canada) attached to the facemask. Gas flows to the manifold are controlled by a computer-controlled gas blender (RespirAct™, Thornhill Research, Toronto, Canada) programmed to provide the flow and proportion of O₂, N₂, and CO₂ needed to attain the target end-tidal partial pressures of CO₂ (PetCO₂) and O₂ (PetO₂). Our target gas sequence is 8 minutes and 20 seconds in duration, with target PetO₂ of 100 mm Hg throughout, and target PetCO₂ alternating between normocapnia (40 mm Hg) and hypercapnia (50 mm Hg) (Figure 4–1). Tidal pCO₂ and pO₂ are monitored continuously (RespirAct™), digitized, and recorded using data acquisition software (LabView, National Instruments Corporation, Austin, TX). The apparatus and technique are described in greater detail elsewhere.^3,4

Figure 4–1 Target gas sequence for PetO₂ (black line) and PetCO₂ (gray waveform).

MRI: Although imaging can be performed at 1.5 Tesla (50% longer data acquisition time to compensate for reduced signal at this field strength), we currently use a 3.0 Tesla system (Signa HDx; GE Healthcare, Milwaukee, WI) with an eight-channel, phased-array head coil. A T1-weighted three-dimensional spoiled gradient echo sequence is used to acquire anatomical images for co-registration (matrix: 256 × 256; slice thickness: 2.2 mm; no interslice gap). During the controlled changes in PetCO₂, BOLD MR acquisitions are obtained. Twenty-six slices are acquired every 2 seconds over a total acquisition time of 8 minutes and 20 seconds. The BOLD sequences is a T2⁎-weighted single-shot gradient echo echo-planar acquisition (field of view: 24 × 24 cm; matrix: 64 × 64; TR: 2000 ms; TE: 30 ms; flip angle: 85 degrees; slice thickness: 5.0 mm; no interslice gap; number of frames: 254).

Data Processing: MRI and PetCO₂ data are imported into AFNI,²² a functional neuroimaging software package available for download at http://afni.nimh.nih.gov/afni/. Total MR signal is calculated for each whole-brain acquisition (i.e., for every set of images covering the entire brain), yielding a whole-brain BOLD MR signal waveform of 500 seconds duration. This whole-brain BOLD MR signal waveform is temporally shifted to the point of maximum statistical correlation with the PetCO₂ waveform (output from the RespirAct). Figure 4–2 shows PetCO₂ (in red) and whole-brain BOLD MR signal (in blue) as a function of time for a normal subject. Note how each increase in PetCO₂ is accompanied by a corresponding increase in BOLD MR signal.

Figure 4–2 PetCO₂ (in red) and whole-brain BOLD MR signal (in blue) as a function of time for a normal subject. This was an 8-minute, 20-second CVR acquisition.

The BOLD MRI signal time course is then regressed (least squares) against the PetCO₂ waveform on a voxel-by-voxel basis. The slope of the regression, expressed as %ΔBOLD MR signal per mm Hg ΔPetCO₂, is CVR. The CVR map is color-coded using a two-color continuous spectrum ranging from red for positive CVR to blue for negative CVR, and then overlaid voxel by voxel on the anatomical images (Figure 4–3).

Figure 4–3 BOLD CO₂ CVR map (%ΔBOLD MR signal per mm Hg ΔPetCO₂) overlaid on T1-weighted anatomical images in a young healthy subject.

BOLD MR measurement of CVR in the EC-IC bypass population

We routinely use BOLD CO₂ CVR to evaluate possible candidates for EC-IC bypass. The majority of these patients have Moyamoya disease or Moyamoya syndrome, and a smaller number have intracranial steno-occlusive disease of atherosclerotic or other etiology. We map CVR in patients for whom surgical revascularization is contemplated, but also to follow those patients who are not surgical candidates, and to postoperatively evaluate those who undergo bypass. The following cases illustrate the technique:

• Case 1: Paradoxical reactivity on BOLD CO₂ CVR (Figure 4-4 through 4-6)

• Case 2: Bilateral disease and BOLD CO₂ CVR evidence of bypass failure (Figure 4-7 through 4-9)

• Case 3: Quantitative BOLD CO₂ CVR (Figure 4–10)

Figure 4–4 Twenty-three-year-old man with neurofibromatosis type 1 and secondary Moyamoya syndrome. Contrast-enhanced MR angiogram shows stenosis of the terminal right ICA and right MCA, and Moyamoya-type collateral vessels about the circle of Willis bilaterally.

Figure 4–5 Initial BOLD CO₂ CVR study (top row) demonstrates paradoxical reactivity in the right anterior and middle cerebral artery territories. The patient underwent right-sided superficial temporal artery (STA) to MCA bypass 3 months later, and a follow-up CVR study 5 months postbypass (bottom row) demonstrates near normalization of CVR. Units are %ΔBOLD MR signal per mm Hg ΔPetCO₂.

Figure 4–6 A plot of BOLD MR signal (blue) versus PetCO₂ (red) for a single voxel with paradoxical reactivity shows that each increase in PetCO₂ is accompanied by a decrease in BOLD MR signal.

Figure 4–7 Thirty-eight–year-old women with Moyamoya syndrome. Time-of-flight MR angiogram shows severe stenosis of the distal ICAs and proximal MCA M1 segments.

Figure 4–8 Initial BOLD CO₂ CVR study (top row) demonstrates paradoxical reactivity in the ACA and MCA territories bilaterally with preservation of CVR in the posterior cerebral artery (PCA) territories. The patient underwent left-sided STA-MCA bypass 2 months later, and 1 month following that, right-sided direct STA-MCA bypass in addition to right-sided indirect bypass. A follow-up CVR study (bottom row) 6 months postsurgery demonstrates moderate improvement in CVR in the anterior circulation bilaterally, although somewhat less than is typical following bypass. The right side has improved more than the left. Units are %ΔBOLD MR signal per mm Hg ΔPetCO₂.

Figure 4–9 Maximum intensity projections from 3D time-of-flight MR angiogram show stenosis at the STA (white arrows) to MCA (black arrows) anastomoses bilaterally, explaining the moderately improved CVR following bypass.

Figure 4–10 Twenty-two-year-old woman with Moyamoya disease involving the anterior circulation bilaterally. A BOLD CO₂ CVR study was performed (top row), the patient underwent left-sided STA-MCA bypass 14 weeks later, and repeat BOLD CO₂ CVR study (bottom row) was performed 13 weeks following bypass. CVR in the MCA gray matter ipsilateral to EC-IC bypass increased from 0.02% ΔBOLD MR signal per mm Hg ΔPetCO₂ before bypass to 0.25% after bypass. The contralateral side demonstrated a milder increase from 0.12% ΔBOLD MR signal per mm Hg ΔPetCO₂ before bypass to 0.21% after bypass.

Qualitatively, BOLD CO₂ CVR maps convey important information. They demarcate a clear difference between positive CVR and paradoxical CVR, and we know that paradoxical CVR indicates significant hemodynamic compromise.

The term “quantitative” can have many meanings in the context of CVR. Some groups conducting CVR studies use the contralateral hemisphere as an internal standard and report “quantitative” interhemispheric asymmetry indices, typically a ratio of right-hemisphere CVR to left-hemisphere CVR. Other groups have used cerebellar CVR as an internal standard. The disadvantage of these approaches is that they are only valid in patients with unilateral disease or lack of posterior circulation involvement, respectively. This is a major limitation. The major advantage of our approach is that it is independent of the number and degree of vascular stenoses.

In addition to patient factors (CBV, CMRO₂, pO₂, hematocrit) discussed earlier, BOLD MR signal depends on both MR hardware (most notably magnetic field strength) and MR pulse sequence specifics (sequence type, echo time, voxel size, etc.). When performing CVR studies using one MR scanner with consistent sequence parameters, CVR measured as %ΔBOLD MR signal per mm Hg ΔPetCO₂ is highly reproducible. This does not provide a measure of CVR in the gold-standard CBF units of ml/100 g/min, but it does enable quantitative comparisons within individual patients over time, and among patients. The ability to quantitatively compare BOLD CO₂ CVR values across institutes using different MR scanners and MR sequences will require further development, but interinstitute reproducibility studies^23,24 of myocardial T2⁎ measurements suggest that this is feasible.

We are currently studying the ability of preoperative CVR to predict the degree of hemodynamic improvement post-bypass. To date we have studied 16 consecutive patients with arterial steno-occlusive disease undergoing EC-IC bypass surgery. Preliminary results²⁵ show that ipsilateral MCA territory BOLD CO₂ CVR (%ΔBOLD MR signal per mm Hg ΔPetCO₂) improves significantly following revascularization (0.07% ± 0.08% prebypass to 0.24% ± 0.07% postbypass [two-sided p < 0.01]). Greater impairment of CVR preoperatively correlates with greater improvement in that territory following bypass (r = 0.73, p < 0.01). Figure 4–10 illustrates a case from this series.

II DSC MR Measurement of CVR

Whereas BOLD MRI exploits the magnetic properties of endogenous deoxyhemoglobin, dynamic susceptibility contrast (DSC) MRI is based on imaging following intravenous injection of a gadolinium chelate. Both deoxyhemoglobin and gadolinium possess a property called “magnetic susceptibility” that enables their detection on MRI, and this is reflected in the name of this technique. Both operate on the same principle of contained or bounded paramagnetic substances reducing the MRI signal. In the case of DSC, it is gadolinium that is contained by the walls of the microvessels. Methods of DSC MRI measurement of cerebral perfusion have been described in detail elsewhere. A typical approach follows.

A T2⁎-weighted echo-planar MR pulse sequence (which is sensitive to the susceptibility effects of gadolinium) is used to repeatedly image the whole brain every 1 to 2 seconds. This records the baseline signal intensity of each voxel prior to the gadolinium bolus. After several whole-brain acquisitions, a tight bolus of intravenous contrast is infused using a power injection of gadolinium chelate at 5 ml/s through an 18-gauge intravenous catheter in the antecubital fossa. Brain imaging continues during and after contrast injection, for a total of 90 seconds, capturing the first pass of gadolinium through the microcirculation. It has been shown that the signal change in a voxel is directly related to the amount of gadolinium in that voxel,²⁶ and thus the signal-time curve for each voxel is converted into a tissue concentration–time curve. Calculation of CBF requires one additional parameter: an arterial input function (AIF). The AIF describes the concentration-time curve of the contrast bolus on the arterial side, before it passes through the tissue (capillaries). There are a variety of methods for determining the AIF, but most often it is measured as the concentration-time curve for a single region of interest drawn near a major artery supplying the brain. CBF is then calculated based on indicator-dilution theory for a nondiffusible tracer. There are commercial software packages that perform this postprocessing, generating maps of perfusion including relative blood flow, blood volume, and transit time. The extension of DSC MRI perfusion to DSC MRI CVR simply involves performing DSC MRI before and after a vasodilatory stimulus, and then calculating the percent change in CBF per unit vasodilatory stimulus on a voxel-by-voxel basis.

DSC MRI CVR shares some of the advantages of BOLD CO₂ CVR: wide availability, lack of ionizing radiation, relatively high spatial resolution, and ability to obtain anatomical MRI and MRA in the same imaging session. A potential advantage of DSC MRI over BOLD CO₂ MRI is the ability to measure CVR using the standard units of milliliters per 100 g per minute. However, a major disadvantage compared with BOLD MRI is the requirement for an arterial input function. The choice of a reference artery is critical for accurate CBF measurement; this remains non-straightforward and there is no general consensus on how this should be done. This is a particular challenge in patients with severe steno-occlusive disease, as large artery involvement can preclude use of the arteries typically used for the AIF. It is even more problematic when there are bilateral vascular occlusions. Current research in this area is focused on deriving input functions at the tissue level rather than at large feeding arteries.

There is a wealth of literature on DSC for measuring perfusion, but only a few reports of DSC CVR. Schreiber et al.²⁷ studied 13 normal subjects and eight patients with steno-occlusive disease using DSC acetazolamide CVR and found reduced CVR in disease patients compared with normals. Similarly, Guckel et al.²⁸ studied 21 symptomatic patients with occlusive cerebrovascular disease using DSC acetazolamide CVR, and demonstrated significantly reduced CVR in the affected hemisphere compared with the contralateral side. Ma et al.²⁹ studied 12 patients with symptomatic unilateral ICA occlusion using DSC acetazolamide CVR and Tc-99m ECD SPECT acetazolamide CVR in each patient. Statistical conclusions were limited given the small sample size: seven of nine patients with impaired CVR on SPECT had impaired CVR on DSC MRI, and two of three patients with normal CVR on SPECT had normal CVR on DSC.

III ASL MR Mapping of CVR

In its most basic form, ASL MRI uses radiofrequency pulses to magnetically label water protons that then flow into a selected imaging slice in the brain. These modified protons then act as an endogenous contrast agent causing an MRI signal change that is directly proportional to CBF. There are many technical variations on this basic approach.

ASL MRI CVR shares some of the advantages of BOLD CO₂ CVR: lack of ionizing radiation, no need for intravenous injection, and ability to obtain anatomical MRI and MRA in the same imaging session. A potential advantage over BOLD CO₂ CVR is the ability to quantify CVR in standard units; however, quantitative ASL analysis is complex and currently not widely available for clinical use. A major disadvantage of ASL is lower signal-noise ratio than BOLD. Another disadvantage is that delayed arterial transit can cause a significant overestimation of the CVR deficit due to loss of the water proton label through T1 relaxation mechanisms.

ASL MRI CVR in a cerebrovascular disease population has been demonstrated by Detre et al.,³⁰ Arbab et al.,³¹ and our group,¹⁸ but further validation and dissemination of the MR pulse sequences are still needed before this can become a routine clinical technique.