Blood Oxygen Level–Dependent MRI of Cerebral CO2 Reactivity in Severe Carotid Stenosis and Occlusion
Background and Purpose— Impaired cerebrovascular reserve capacity (CVC) is a risk factor for ischemic events in patients with high-grade carotid stenosis and occlusion. In this study, the CVC in response to a CO2 challenge was evaluated with blood oxygen level–dependent (BOLD) MRI and the results compared with those of a transcranial Doppler CO2 tests.
Methods— A T2*-weighted single-shot multigradient echo-planar imaging sequence was used to determine cerebral CO2 reactivity. T2* values were calculated for each pixel at rest and during a challenge with 7% CO2, and a reference function was fitted to the T2* time courses. Whole-brain color-coded ΔT2* parameter maps were calculated and visually evaluated for regional differences. Additionally, a region-of-interest analysis was undertaken. Average values for ΔT2* normalized to changes in end-tidal Pco2 were calculated. Results were correlated with a transcranial Doppler CO2 tests in 20 patients with high-grade stenosis or occlusion of the carotid artery.
Results— Color parameter maps showed areas of decreased BOLD effect within the internal carotid artery territory in 12 of 13 hemispheres with impaired CVC in transcranial Doppler CO2 test. Regional normalized ΔT2* was highly correlated with changes of middle cerebral artery blood flow velocity in transcranial Doppler CO2 test. Normalized ΔT2* was significantly reduced in hemispheres with impaired CVC in transcranial Doppler (P<0.0001).
Conclusions— BOLD MRI can easily be included in routine MRI exams. The technique is robust and yields diagnostic information concerning the cerebrovascular reserve.
High-grade stenosis or occlusion of the carotid artery (CA) may reduce perfusion pressure in the dependent brain territory when collateral flow is insufficient. Consequently, autoregulatory vasodilatation occurs to maintain regional cerebral blood flow (CBF) within normal limits. This results in decreased or even exhausted cerebrovascular reserve capacity (CVC). CVC is determined by measurements of CBF at rest and after exposure to vasodilatatory stimuli, such as inhalation of CO2, intravenous administration of acetazolamide, and breath holding. Decreased CVC is an independent risk factor for ischemic events in patients with carotid stenosis and occlusion.1,2
Blood oxygen level–dependent (BOLD) contrast MRI relies on changes in blood oxygen saturation during repetitive measurements. Deoxyhemoglobin is paramagnetic, and oxyhemoglobin is diamagnetic. The concentration of deoxyhemoglobin consequently affects the magnetic susceptibility properties of blood and the transverse relaxation rate T2* of the surrounding tissue.3 BOLD contrast is applied widely in functional MRI.4,5 In activated brain areas, the increase in regional CBF exceeds oxygen demands, raising blood oxygen saturation and signal intensity on T2*-weighted images. It has been shown that BOLD MRI can also be applied to CVC examination.6 BOLD MRI is completely noninvasive, not even requiring administration of intravenous contrast media. However, the clinical usefulness of the method has not been demonstrated until now. In this article, quantitative full-brain BOLD MRI of CVC with optimized data acquisition and processing is evaluated. The clinical utility is determined in comparison with transcranial Doppler (TCD) CO2 testing.
Subjects and Methods
A total of 27 consecutive patients (11 women and 16 men) with unilateral or bilateral severe steno-occlusive disease of the CA were studied between April 2003 and February 2004. Mean age was 66 years (range 42 to 82).
CA disease was graded by using Doppler frequency shifts prestenotically, intrastenotically, and poststenotically in combination with B-mode imaging.7 Details about the patients and their disease(s) are given in the Table. Symptomatic patients were studied at least 1 week after the last ischemic event.
Magnetic Resonance Protocol
MRI was performed on a 1.5T scanner (Sonata; Siemens). A transversal T2-weighted fluid-attenuated inversion recovery sequence and a transversal diffusion-weighted sequence were acquired for detection of chronic and acute infarction, respectively. In addition, an arterial 3D time-of-flight magnetic resonance angiography (MRA) of the circle of Willis was acquired and reconstructed as maximum intensity projections. BOLD imaging was performed with a single-shot multigradient echo-planar imaging (EPI) sequence.8 Twenty slices were acquired in the transversal plane. Repetition time was 3000 ms, flip angle 90°, matrix size 64×64, field of view 220×220 mm, and slice thickness 5 mm. Four echo images were read out per measurement, with echo times of 17, 44, 71, and 98 ms, respectively. A total of 100 measurements were acquired. During measurements 21 through 60 (ie, during 2 minutes), room air enriched with 7% CO2 was administered via a breathing mask covering the nose.
Patient monitoring comprised continuous ECG recording, pulse oximetry, and recording of inspiratory, end-tidal Pco2 (PETCO2), and blood pressure (MAGLIFE C; Schiller).
In addition to patient examinations, 5 healthy volunteers (3 men and 2 women; mean age 30 years) were studied with the same protocol. In these studies, 3 different inspiratory CO2 concentrations of 3%, 5%, and 7% CO2 were administered.
Quantitative I0 and T2* values were determined pixel-wise using a monoexponential model. The maps were motion-corrected during the reconstruction process.9 The T2* maps were spatially filtered with a 3D Gaussian filter (full width at half maximum 10 mm) and the T2* time courses were filtered with a median filter (kernel size 5).
A pixel-wise least square data fit of a model function to the T2* time course was performed. The reference function is based on a compartment model of CO2 diffusion between the pulmonary air spaces and blood along the alveolar membrane and is mathematically described as equation
where y1=baseline T2* before and y2=baseline T2* after administration of CO2, τ1=time constant of exponential increase and τ2=time constant of exponential decrease of T2*, t1=start and t2=end of CO2 administration, ydiff=mathematical maximum increase of the e-function. ΔT2*=T2*max−(y1+y2)/2.
The rationale for the data fit is to eliminate the oscillations of T2* to define a baseline and maximum. The T2* time course is shown for an individual pixel together with the fitting result in Figure 1. The variables ΔT2*=T2*max−T2*base, ΔR2*=1/ T2*base−1/ T2*max, and relative ΔT2*=(T2*max/T2*base)−1 were calculated from the fitted data and displayed as parameter color maps. These parameter maps were evaluated by visual inspection by an experienced neuroradiologist (S.Z.).
Regions of interest (ROIs) for the right and left middle cerebral artery (MCA) territory comprising all slices without susceptibility artifacts from the skull base were manually defined on T2*-weighted images. ROI-based averages of ΔT2* were calculated and normalized to the change in PETCO2 (ΔPETCO2) for better comparability with TCD CO2 testing. The results of these ROI analyses were included in the statistical evaluation.
For statistical analysis, each particular hemisphere was classified as normal or decreased in CVC according to the results of TCD CO2 testing. Normal distribution was proven and a 2-tailed t test for normalized ΔT2* was performed. Pearson correlation coefficients for normalized ΔT2* and the change in normalized mean CBF velocity (CBFV) as determined in TCD CO2 testing (ΔCBFV/ΔPETCO2) were calculated. SPSS software was used for statistics.
TCD CO2 Testing
Measurements were performed with subjects in a supine position. Both MCAs were insonated through the temporal bone window using 2-MHz transducers attached to a headband (DWL-Multidop-X4; Sipplingen). PETCO2 was measured continuously with an infrared capnometer (Normocap; Datex). Baseline flow velocity was measured for 1 minute. Subsequently, a 7% CO2–air mixture was administered until reaching stable hypercapnia, which was maintained for 60 to 90 s. The measurements were continued for 60 s until baseline values were reached after withdrawal of CO2. The CO2 reactivity was calculated as the maximum percentage increase in mean CBFV during hypercapnia divided by the increase in PETCO2 in kilopascals. The determination of CO2 reactivity in the nonstenosed or minor stenosed sides of 127 patients (66±9 years) with unilateral severe carotid stenosis yielded a lower threshold of 8.25%/kPa (5% quantile). This value was defined as the cutoff to clearly pathological CO2 reactivity.
BOLD MRI and TCD CO2 testing were performed within several hours to 7 days in a random sequence. Operators were blinded to the results of the other method. Results of TCD CO2 testing were unknown when the parameter maps were evaluated by visual inspection. The study was approved by the local ethics committee, and all participants gave written informed consent.
Volunteer studies indicated a dependence of ΔT2* from inspiratory CO2 concentration (Figure 2). Color parameter maps showed the highest BOLD response in cortical and deep gray matter and within venous vessels with a maximum ΔT2* of ≈10 to 12 ms. However, a less pronounced effect could also be clearly demonstrated in white matter. No significant differences between parameter maps of ΔT2*, relative ΔT2*, and ΔR2* were observed.
In the patient group, average ΔPETCO2 was 1.75 kPa (SD 0.4). Mean heart rate and O2 saturation were 67.1/min (SD 13.2) and 94.8% (SD 1.5) during normocapnia, and 71.8/min (SD 16.9) and 95.4% (SD 1.6) during hypercapnia. The hypercapnia-induced average increase in mean arterial blood pressure was 13 mm Hg (1.74 kPa).
Seven patients, all with unilateral high-grade stenosis, could not be evaluated by TCD CO2 testing because of an insufficient bone window.
A total of 40 hemispheres were evaluated in the patient group with sufficient transtemporal ultrasound transmission. Nine patients (18 hemispheres) with bilaterally normal CVC in TCD CO2 testing showed a symmetrically distributed normal BOLD response on color parameter maps. One patient with occlusion of the left CA showed normal CVC in TCD CO2 testing. The BOLD effect was clearly reduced in the left MCA territory on ΔT2* parameter maps.
TCD CO2 testing showed a decreased CVC in 1 hemisphere in 7 patients. In 6 of 7 affected hemispheres, BOLD MRI demonstrated large ipsilateral areas of reduced BOLD response (Figure 3). Visual inspection of the parameter maps did not reveal clear abnormalities in 1 patient with unilateral impairment of CVC in TCD CO2 testing.
Both hemispheres were affected by reduced CVC in TCD CO2 testing in 3 patients with bilateral high-grade stenosis or occlusion. BOLD MRI parameter maps showed a bihemispherically reduced response in the latter patients. However, these maps were difficult to interpret because of their inhomogeneous appearance and the lack of a normal reference hemisphere.
In the patient group, hemispheres with normal CVC in TCD CO2 testing showed an average ΔT2* normalized to ΔPETCO2 of 1.8 ms/kPa (1SD 0.57). Hemispheres with impaired CVC in TCD CO2 testing showed a reduced average normalized ΔT2* of 0.7 ms/kPa (1SD 0.64; Figure 4a). The patient with unilateral impairment of CVC in TCD CO2 testing and normal ΔT2* parameter maps showed a normalized ΔT2* of 1.9 ms/kPa on the affected and 2.1 ms/kPa on the contralateral side. The patient with bilaterally normal CVC in CO2 TCD CO2 testing and reduced BOLD effect in the left MCA territory showed an ipsilateral ΔT2* of 0.8 ms/kPa (contralateral 1.3 ms/kPa).
The difference between hemispheres with normal and hemispheres with impaired CVC in TCD CO2 testing was statistically highly significant (P=0.0001). A linear correlation between normalized ΔT2* and ΔCBFV in TCD CO2 testing (Figure 4b) could be clearly shown, with r=0.71 (P<0.001).
In 2 of 7 patients with absent temporal ultrasound window color, T2* maps showed large areas with reduced BOLD effect in the ipsilateral MCA and internal CA territory, respectively. Quantitative ROI-based analysis revealed a low mean normalized ΔT2* of 1.1 and 0.3 ms/kPa in the ipsilateral MCA territory, and 2.0 and 2.3 ms/kPa on the contralateral side. In 5 of 7 patients, ΔT2* parameter maps did not show any side differences. These patients showed an average normalized ΔT2* of 2.1 (range 1.4 to 4.2 ms/kPa).
Inspection of arterial time-of-flight angiography revealed absent or significantly reduced flow-related signal attributable to slow flow and saturation in the MCA of 8 of 15 hemispheres with reduced and 3 of 39 hemispheres with normal BOLD effect, respectively.
Various methods exist for CVC determination. In xenon-enhanced computed tomography (CT), single-photon emission CT10 and positron-emission tomography,11 CBF is determined before and after administration of a vasodilatatory stimulus. Furthermore, positron-emission tomography may demonstrate increased oxygen extraction fraction and cerebral blood volume in severe hemodynamic impairment.12 These methods are expensive and the subject is exposed to radiation. The completely noninvasive evaluation of CVC with TCD CO2 testing is presumably the most widely applied method in clinical practice. Drawbacks of TCD CO2 testing are an insufficient temporal bone ultrasound window in ≈10% to 20% of the patients and relatively low diagnostic accuracy (index of validity ≈75%, compared with xenon-enhanced CT13).
First attempts to determine CVC by means of BOLD MRI date back to 1994, when changes in R2* after administration of acetazolamide and CO2 were observed in the rat brain14 and CO2-induced signal increase in normal subjects could be demonstrated.6 The effect was conspicuous in cortical and deep gray matter but was not significant in cerebral white matter.6 In 1995, first results of BOLD contrast MRI in 4 patients with unilateral CA occlusion were reported.15 In this study, only a single slice could be acquired and evaluated by means of simple difference images and ROI analysis of pial veins. It could be shown that acetazolamide-induced signal increase in pial veins was lower in patients with reduced ipsilateral CVC. In a larger series of patients with unilateral CA stenosis, Lythgoe et al16 found a significant correlation between the side difference of hemispheric reactivity determined by BOLD MRI and TCD CO2 testing, respectively. However, a significant correlation between the particular hemispheric relative BOLD signal increase and CO2 reactivity in TCD CO2 testing could not be demonstrated.
In contrast to the study by Lythgoe et al,16 our study demonstrated a significant correlation between normalized ΔT2* and ΔCBFV/ΔPETCO2 in TCD CO2 testing. This is probably attributable in part to the enhanced sensitivity of single-shot multigradient-echo functional MRI compared with conventional EPI sequences.9,17,18
Because BOLD effects are small, they can best be depicted by statistical analysis of repeated measurements. The greater amount of acquired data may also have enhanced the sensitivity of our method. The number of measurements in our study was 5× higher than for Lythgoe et al.16 By the acquisition of 4 different echoes, our sequence further increases the number of acquired data points by a factor of 4.
As another advantage, the method allows a calculation of parameters with defined dimensions, such as ΔT2* in milliseconds instead of dimensionless signal intensities. Consequently, quantitative results and comparisons can be achieved. It could be demonstrated in healthy subjects that ΔCBF can be calculated if baseline T2* and ΔT2* are available.19 This potential further development of our method was not included in our article because the assumptions regarding the relation of blood volume and flow (Grubb’s exponent) are not proven to be true in our patient population.
The fit of the T2* time course is advantageous because it is insensitive to differences in the temporal evolution of the vascular response to CO2. It could be shown that the time course of the BOLD signal change in response to CO2 differs between white and gray matter. The BOLD signal rise in white matter is much slower and sustained.20 Because the white matter comprises approximately one third of the brain volume,21 its contribution to hemispheric BOLD effect should not be neglected.
An impaired CVC may be diagnosed at first glance by use of ΔT2* parameter maps if 1 hemisphere shows an unambiguously reduced BOLD response. A quantitative ROI-based analysis appears valuable in cases with bilaterally impaired CVC, and it may increase the sensitivity and specificity of BOLD MRI in cases with ambiguous results on solely visual inspection. However, a larger number of patients, as well as normal controls, must be examined to establish normal values.
Inspection of ΔT2* parameter maps and quantitative ROI-based analysis were in good agreement with TCD CO2 testing in most cases. Results were discrepant in 2 cases. One patient with unilateral asymptomatic CA occlusion showed a reduced CVC in TCD CO2 testing but normal ΔT2* parameter maps. This might reflect the presence of sufficient blood supply via leptomeningeal collateral arteries. In 1 patient with unilateral asymptomatic CA occlusion, TCD CO2 testing was normal, but BOLD MRI showed a pathological CVC. Because the flow-related vascular signal of the ipsilateral MCA was severely reduced, it may be speculated that results of TCD CO2 testing were false-negative in this case.
Time-of-flight MRA may, to some degree, anticipate results of BOLD MRI. A severely saturated flow signal indicates a higher probability of impaired CVC. However, there are patients with only slight reduction of flow-related signal in the ipsilateral MCA and reduced CVC. On the other hand, hemispheres with severely reduced or absent flow-related signal can show normal CVC. The latter constellation may be suggestive of sufficient leptomeningeal collateral flow.
The method presented in this article is mature and robust. It can easily be included in a routine MRI examination and may serve as an alternative to xenon-enhanced CT or nuclear medicine methods in cases in which TCD CO2 testing cannot be successfully applied. BOLD-based CVC measurements in combination with TCD CO2 testing may even offer additional information otherwise unavailable: a reduced CVC in TCD CO2 testing together with a normal CVC in BOLD MRI may be indicative of sufficient collateral blood supply to the relevant vascular territory.
CVC can be determined by BOLD MRI and administration of CO2. BOLD measurements can easily be included in clinical MRI sessions, representing an alternative to TCD CO2 testing, which is sometimes hampered by an insufficient bone window.
S.Z. contributed conceptual study design, execution of the study, and writing of this manuscript; J.R. was responsible for development of data evaluation with the data fit; M.R. contributed evaluation of transcranial Doppler CO2 tests; A.H. was responsible for conceptual study design and evaluation of transcranial Doppler CO2 tests; I.M. contributed data analysis and statistics; and O.S. performed development of data evaluation with the data fit, programming of the multiecho gradient-echo sequence, and supervision of data acquisition.
- Received July 16, 2004.
- Revision received December 3, 2004.
- Accepted December 21, 2004.
Markus H, Cullinane M. Severely impaired cerebrovascular reactivity predicts stroke and TIA risk in patients with carotid artery stenosis and occlusion. Brain. 2001; 124: 457–467.
Ogawa S, Lee TM, Kay AR, Tank DW. Brain magnetic resonance imaging with contrast dependent of blood oxygenation. Proc Natl Acad Sci U S A. 1990; 87: 9868–9872.
Kwong KK, Belliveau JW, Chesler DA, Goldberg IE, Weisskoff RM, Poncelet BP, Kennedy DN, Hoppel BE, Cohen MS, Turner R, Cheng H, Brady TJ, Rosen BR. Dynamic magnetic resonance imaging of human brain activity during primary sensory stimulation. Proc Natl Acad Sci U S A. 1992; 89: 5675–5679.
Latchaw RE, Yonas H, Hunter GJ, Yuh WTC, Ueda T, Sorensen G, Sunshine JL, Biller J, Wechsler L, Higashida R, Hademenos G. Guidelines and recommendations for perfusion imaging in cerebral ischemia. Stroke. 2003; 34: 1084–1104.
Kuwabara Y, Ichiya Y, Sasaki M, Yoshida T, Fukumura T, Masuda K, Fujii K, Fukui M. PET evaluation of cerebral hemodynamics in occlusive cerebrovascular disease pre- and postsurgery. J Nucl Med. 1998; 39: 760–765.
Derdeyn CP, Videen TO, Yundt KD, Fritsch SM, Carpenter DA, Grubb RL, Powers WJ. Variability of cerebral blood volume and oxygen extraction: stages of cerebral hemodynamic impairment revisited. Brain. 2002; 125: 595–607.
Pindzola RR, Balzer JR, Nemoto EM, Goldstein S, Yonas H. Cerebrovascular reserve in patients with carotid occlusive disease assessed by stable Xenon-enhanced CT cerebral blood flow and transcranial Doppler. Stroke. 2001; 32: 1811–1817.
Kleinschmidt A, Steinmetz H, Sitzer M, Merboldt KD, Frahm J. Magnetic resonance imaging of regional cerebral blood oxygenation changes under acetazolamide in carotid occlusive disease. Stroke. 1995; 26: 106–110.
Speck O, Günther S, Kiselev V, Bilecen D, Bellemann ME, Hennig J. Quantitative determination of the perfusion reserve capacity using single shot T2* mapping during CO2 challenge. Proceedings of the 10th Annual Meeting of the ISMRM, Honolulu, Hawaii, April 18–24, 2002; 264.