T2′ Imaging Within Perfusion-Restricted Tissue in High-Grade Occlusive Carotid Disease
Background and Purpose—Quantitative T2′ imaging presumably detects regional changes in the relation of oxygenated and deoxygenated hemoglobin. Regional differences in hemoglobin oxygenation might reflect areas with increased oxygen extraction for compensation of reduced perfusion pressure. We investigated quantitative T2′ imaging in patients with high-grade stenoses of brain-supplying arteries and hypothesized that T2′ values are lower in perfusion-restricted areas as compared with normally perfused tissue.
Methods—Eighteen patients (15 men; mean age±SD, 54±12.8 years) with unilateral symptomatic or asymptomatic high-grade extracranial or intracranial internal carotid artery or proximal middle cerebral artery stenosis/occlusion were included. MR examination included perfusion-weighted imaging and quantitative, motion-corrected mapping of T2′ time. Time-to-peak and mean transit time maps were thresholded for different degrees of perfusion delays (eg, >0 seconds, ≥2 seconds) compared with the contralateral hemisphere. Mean T2′ values in areas of impaired perfusion were compared with T2′ values in corresponding contralateral or ipsilateral, normoperfused areas.
Results—Mean size of perfusion-impaired areas in time-to-peak maps (time-to-peak delay >0 seconds) was 10.8 mL (±6.3) and 11.5 mL (±6.4) in mean transit time maps (mean transit time delay >0 seconds). T2′ values were significantly (P<0.01) lower in all perfusion-restricted compared with corresponding contralateral brain areas (ipsilateral versus contralateral). For time-to-peak delay >0 seconds, T2′ values were 115 ms (±9) versus 125 ms (±12). For mean transit time delay >0 seconds, T2′ values were 115 ms (±9) versus 128 ms (±10). Differences in T2′ values increased with the severity of the perfusion delay. Ipsilateral T2′ values outside the perfusion-disturbed areas did not differ from contralateral T2′ values.
Conclusions—Motion-corrected T2′ imaging presumably detects areas with increased oxygen extraction within perfusion-restricted tissue in patients with high-grade occlusive vessel disease.
Atherosclerotic large-vessel stenosis is the leading cause for ischemic stroke worldwide.1 Besides the degree of the stenosis, the extent of perfusion disturbances determines the risk of subsequent cerebral ischemia. However, the association between both parameters is weak, mainly due to large interindividual variation in collateral vessel status.2,3
Using positron emission tomography, Baron et al4 developed a 2-stage model of autoregulatory mechanisms after disturbances of blood circulation. This model was further modified by others.2,5 In the case of a decreased perfusion pressure, blood vessels dilate, whereas the oxygen extraction fraction (OEF) remains relatively stable. When vasodilatation capacity is exhausted, further decrease of perfusion pressure leads to an increase of OEF in an attempt to satisfy metabolic demands of brain tissue.
Quantitative T2′ MRI has the potential to detect differences in the blood oxygen saturation using the blood oxygen level-dependent effect and might therefore be a surrogate parameter of OEF.6 Although oxygenated hemoglobin is diamagnetic, deoxygenated hemoglobin is paramagnetic leading to a more profound signal decay on T2*-weighted images. For T2′ mapping, T2*-weighted images are corrected for T2 relaxation effects. T2′ maps may provide reliable information on the extent of blood oxygenation irrespective of signal alterations caused by gliosis or edema.
The capability of T2′ imaging to predict penumbral tissue in acute stroke has been shown previously, yielding additional information to conventional perfusion- and diffusion-weighted imaging.7,8 In chronic cerebral hypoperfusion caused by high-grade stenosis, T2′ mapping might also be valuable to delineate areas with increased O2 consumption due to impaired perfusion. In this proof-of-concept study, we hypothesized that T2′ values are reduced in areas with substantial perfusion disturbances in patients with high-grade occlusive carotid disease.
Materials and Methods
Eighteen consecutive patients (15 men; mean age±SD, 54±12.8 years; range, 28–75 years) with unilateral symptomatic or asymptomatic high-grade stenosis or occlusion of the internal carotid artery (ICA) or the middle cerebral artery (MCA) were included. Inclusion criteria were (1) Doppler/ultrasound evidence of unilateral, >70% (North American Symptomatic Carotid Endarterectomy Trial [NASCET] criteria) high-grade extracranial ICA stenosis or ICA occlusion; or (2) Doppler/ultrasound or MR angiographic evidence of a high-grade (>50%) unilateral intracranial ICA or proximal MCA stenosis/occlusion. Exclusion criteria were (1) age <18 years; (2) Doppler/ultrasound or MR angiographic evidence of additional contralateral >50% ICA or MCA stenosis, renal insufficiency, MR contraindications and others (ie, psychomotoric restlessness, severe aphasia).
Sites of stenosis/occlusion were: extracranial ICA (n=6), intracranial ICA (n=4), and proximal MCA (M1 segment; n=8). Underlying pathology for stenosis/occlusion was atherosclerosis (n=14), dissection (n=2), central nervous system vasculitis (n=1), and tumoral compression (n=1). Nine patients had a symptomatic (transient ischemic attack or stroke within the past 30 days), and 9 patients had asymptomatic stenosis/occlusion. In 2 patients, contralateral vessels were affected by a low-grade (<50%) stenosis (extracranial ICA: n=1; intracranial MCA: n=1). The study was approved by the local ethics committee and written informed consent was obtained from each patient before study enrollment.
MR measurements were performed on a 3-T whole-body scanner (Trio, Magnetom Series; Siemens, Erlangen/Germany) using the scanner's body coil for radiofrequency transmission and an 8-element phased-array head coil for signal reception. In each patient the MR examination included quantitative imaging (T2 mapping and T2* mapping sequences) as well as diffusion- and perfusion-weighted-imaging, MR angiography, and conventional imaging (T1-weighted, T2-weighted). Four patients were diffusion-weighted imaging-positive with very small diffusion-weighted imaging lesion volumes (<10 mL) mainly within the parietooccipital white matter.
T2 mapping was based on a fast spin echo sequence with an echo train length of 11 echoes per excitation, an echo spacing of 17.1 ms, and the following imaging parameters: 50 axial slices with 2-mm slice thickness, no interslice gap, TR: 10 seconds, bandwidth: 100 Hz/pixel, 180° refocusing pulses, matrix size 192×132 (readout×phase encoding), field of view (FOV) 240×165 mm2, and in-plane resolution 1.25×1.25 mm2. For quantitative T2 mapping, 5 data sets were acquired with different TE values (17, 86, 103, 120, 188 ms), keeping all other acquisition parameters constant. The total duration was 11 minutes 50 seconds.
For quantitative T2* mapping, a series of T2*-weighted images with increasing TE was acquired using a fast low angle shot9 sequence with the acquisition of 8 gradient echoes with identical phase encoding per excitation. Echo formation was achieved by successively inverting the readout gradient. All echoes were acquired under the same readout gradient polarity to avoid misregistrations due to different distortions in the presence of static magnetic field inhomogeneities. Imaging parameters were: linear TE increase from 10 ms to 52 ms with a constant increment of 6 ms, TR: 3000 ms, 50 axial slices with 2-mm slice thickness, no interslice gap, bandwidth: 300 Hz/pixel, flip angle: 30°, matrix size 160×128 (readout×phase encoding), field of view 200×160 mm2, in-plane resolution 1.3×1.3 mm2, and measurement time 6 minutes 24 seconds. Due to the requirement of performing gradient echo imaging with relatively long TE, T2* mapping is very sensitive to motion artifacts, resulting in erroneous T2* values. To avoid this effect, the acquisition was repeated with the same field of view but reducing the spatial resolution by a factor of 2 in phase encoding direction (ie, with 64 phase-encoding lines). Due to the reduction, the acquisition time of this additional experiment amounted to 3 minutes 12 seconds only. After suitable phase corrections to allow for a direct combination of the respective k-space data from both experiments, the full resolution and the reduced resolution data sets were averaged using an individual weighting factor for each k-space line that was chosen in a way to suppress motion affected data.10
Perfusion-weighted imaging was based on a gradient echo echoplanar imaging sequence with the following imaging parameters: TE: 35 ms; TR: 1500 ms; flip angle: 90°; field of view: 192×192 mm; matrix: 64×64; slice thickness: 4 mm; number of slices: 16; voxel size: 2.0×2.0×4.0 mm3; and acquisition time: 75 seconds. The contrast agent (0.1 mmol/kg Gd-DTPA; Magnevist, Schering) was injected into an antecubital vein using a power injector at a rate of 5 mL/s followed by a flush with 10 mL saline.
For verification of intracranial stenosis and exclusion of significant (>50%) contralateral intracranial ICA or MCA stenosis, a 3-dimensional time-of-flight angiography was performed.
Processing of MRI Raw Images
Raw images (turbo spin echo and perfusion-weighted) were linearly coregistered to the first raw T2*-weighted image using the FMRIB Linear Registration Tool (FLIRT–part of FSL, http://www.fmrib.ox.ac.uk/). The quantitative maps were generated with custom-built programs written in MATLAB (MATLAB; Mathworks Inc, Natick, MA). In particular, T2* and T2 relaxation times were obtained pixelwise by exponential fitting of the respective TE dependence of the image signal. Maps of T2′ were calculated as 1/T2′=1/T2*−1/T2. For details, see the online-only Data Supplement.
Perfusion-weighted MRI raw images were processed on a pixel-by-pixel basis to generate maps of the time to peak (TTP) and relative mean transit time (MTT). We determined the shape of the arterial input function in each patient by manually choosing 5 to 10 pixels over the first 2 segments of the MCA in the deep gray matter of the unaffected hemisphere, showing an early and large decrease in signal intensity after the contrast agent injection. For calculation of MTT maps we used the model-independent (nonparametric) singular value decomposition deconvolution method described by Ostergaard et al.11,12
Perfusion maps served as a blueprint to delineate areas with perfusion-disturbed tissue. These areas were analyzed for changes of T2′ as compared with normal perfused tissue.
Region of Interest Placement and Analysis
Two major analysis procedures were performed. In the first analysis, T2′ values in 6 anatomically predefined regions of interest (ROIs) were assessed in both the affected and unaffected hemisphere, irrespective of the presence or absence of perfusion abnormalities. In the second analysis, T2′ values within the areas with perfusion delay (outlined on TTP and MTT maps) were determined and compared with T2′ values in contralateral corresponding areas.
In the first analysis, 12 ROIs (6 in each hemisphere) were placed on T2 maps and transferred to the coregistered and motion-corrected T2′ maps as reported earlier by others.13,14 A standardized circular ROI (approximately 57 mm2) was used, except in the caudate nucleus, where the ROI size had to be adapted individually according to its size and visibility. The ROIs were placed in the frontal white matter, the caput nuclei caudati, the lentiform nucleus, the thalamus, the occipital white matter, and in the semioval center (Figure 1). Due to partial volume effects (cerebrospinal fluid), no ROIs were placed in cortical gray matter. In each hemisphere, 3 ROIs covered white matter (frontal white matter, occipital white matter, semioval center) and 3 ROIs covered deep gray nuclei (caput nuclei caudati, lentiform nucleus, thalamus). Furthermore, ROIs covered different vascular territories (and not exclusively the MCA territory). Mean T2′ values were extracted from each ROI without a new signal fit.
The second analysis compared T2′ values in perfusion-impaired areas with corresponding areas of the contralateral hemisphere. The affected areas were manually outlined on thresholded TTP and MTT maps after determining the extent of perfusion delay in relationship to the unaffected contralateral MCA territory. For details, see the online-only Data Supplement.
For comparison of T2′ values in corresponding ROIs (affected/unaffected hemisphere), the nonparametric Wilcoxon signed rank test was used. Significance level was set at P<0.05. Statistical analysis was performed using SPSS 19 (IBM SPSS Statistics, SPSS Inc).
T2′ Values in Predefined Regions
In the ROI analysis of anatomically predefined regions, no significant differences in T2′ values between the hemisphere with perfusion delay and the normoperfused hemisphere were found (Table 1). In the lentiform nucleus we observed a trend toward lower T2′ values in the affected (66 ms) as compared with the unaffected (74 ms) side (P=0.064), presumably due to the fact that the lentiform nucleus is exclusively supplied by MCA branches, whereas the other regions examined are not. As expected, T2′ values were generally lower in deep gray matter structures (mean±SD, 88±27 ms) as compared with white matter areas (121±24 ms; Table 1).
T2′ Values in Hypoperfused Areas
The size as well as the severity of perfusion restrictions varied substantially between the subjects. Mean size of all perfusion-impaired areas on TTP maps was 10.8 (±6.3 SD) mL ranging from 0 to 20.7 mL decreasing to a mean volume of 5.2 (±4.7) mL in areas with a TTP delay ≥2 seconds. This large variation is presumably due to different degrees of collateral flow among the patients but also due to different sites of vessel occlusion (ICA versus MCA; see the online-only Data Supplement for details).15,16
T2′ values were generally lower in regions with perfusion delay, even in areas with only slight perfusion abnormalities (TTP delay 0–2 seconds). For the entirety of perfusion-impaired areas (TTP delay >0 seconds), T2′ values were 115 ms versus 125 ms (P<0.01) for the unaffected side translating to a mean T2′ reduction of 10 ms (±13; Table 2). Differences in T2′ values between perfusion-disturbed and corresponding contralateral areas were more pronounced in areas with severe perfusion restriction (TTP delay ≥2 seconds: T2′ difference −19 ms) as compared with regions with moderate perfusion delay (TTP delay 0–2 seconds: T2′ difference −6 ms; Figure 2). This difference did not reach statistical significance presumably due to the small sample size of areas with a TTP delay ≥2 seconds (n=10). The numbers of patients with even more pronounced TTP delays (≥4 seconds: n=4, ≥6 seconds: n=2) were too small to draw meaningful conclusions.
When analyzing T2′ values in MTT maps, results were similar to those obtained in TTP maps. In areas with any MTT increase (>0 seconds), T2′ values were 115 ms versus 128 ms in the corresponding unaffected areas (P=0.001; Table 2; Figure 2).
To exclude that T2′ reductions ipsilateral to high-grade stenosis are an unspecific phenomenon not related to perfusion abnormalities, 2 analyses were performed. First, in the patient (No. 5) with high-grade MCA stenosis but without perfusion abnormality, T2′ values within a large ROI (2800 mm2) in the ipsi- and contralateral MCA territory showed almost identical T2′ values: 125 ms ipsilateral and 128 ms contralateral. Second, in 5 patients with focal perfusion abnormalities, circular ROIs (300 mm2) were placed adjacent to the perfusion-disturbed regions (but still within the MCA territory) and in corresponding contralateral areas. T2′ values did not differ significantly between hemispheres (online-only Data Supplement Table II).
The major finding of this study is that T2′ imaging detects areas with increased deoxygenated hemoglobin levels in TTP- and MTT-delayed regions presumably reflecting focally increased OEF. Differences in T2′ values depend on the severity of perfusion abnormalities. To our knowledge, this is the first study combining motion-corrected T2′ mapping with perfusion imaging in patients with high-grade carotid occlusive disease.
T2′ Values in Predefined Regions
Our finding of lower T2′ values in deep gray matter structures as compared with white matter is most likely due to the deposition of iron or other ferromagnetic substances in the deep gray nuclei.13,14 Keeping in mind that there is no evidence of side-specific differences in iron deposition in the basal ganglia,13 we interpret the lower T2′ values in the lentiform nucleus ipsilateral to the stenosis as a result of an increased deoxygenated hemoglobin concentration due to perfusion disturbances within lenticulostriatal arteries originating from the proximal MCA segment. In the other predefined regions, no relevant side-specific T2′ differences were detected, presumably due to normoperfusion in these areas not being exclusively supplied by the MCA and its lenticulostriatal branches. As demonstrated in previous studies, T2′ values were lower in the lentiform nuclei than in the thalamus, which is explained by the more pronounced iron deposition in the lentiform nucleus.
T2′ Values in Hypoperfused Areas
T2′ values were significantly lower in perfusion-impaired areas as compared with corresponding contralateral and to ipsilateral normoperfused tissue. Furthermore, T2′ values decreased with increasing perfusion delay. In areas with severe perfusion abnormalities, collateral blood supply is lower than in areas with no or only slight perfusion disturbances within the territory of the stenotic/occluded artery. Therefore, the signal decline in T2′ maps in perfusion-disturbed brain regions ipsilateral to the stenosis or occlusion most probably reflects an increased amount of deoxygenated hemoglobin as a result of elevated oxygen extraction. This reduction of T2′ values was even seen in only slightly perfusion-disturbed areas, a finding that at first glance seems to contradict the current 2-stage model of autoregulatory mechanisms after reductions in cerebral perfusion pressure. According to this model, OEF remains relatively stable when cerebral perfusion pressure decreases slightly and increases only in severe cerebral perfusion pressure reductions. However, this 2-stage model was recently challenged by a positron emission tomography study indicating that OEF increases can also occur before autoregulatory mechanisms are exhausted.5 Considerable increases of the OEF up to 18% occur in areas with slight cerebral blood flow reductions being within the autoregulatory range, a finding that suggests that OEF rises before the cerebral vasodilatory capacity is exceeded.5,17 We used TTP and MTT maps rather than cerebral blood volume or cerebral blood flow maps for delineation of perfusion-impaired areas. This was done because TTP and MTT maps are highly sensitive for the detection of perfusion abnormalities. Furthermore, easy-to-generate TTP maps have shown to be reliable in detecting severely perfusion-disturbed tissue in acute stroke.18 A study comparing TTP delay assessed by perfusion-weighted MRI with cerebral blood flow/cerebral blood volume -ratio or OEF assessed by positron emission tomography in 24 patients with chronic unilateral carotid occlusive disease found a significant positive correlation between the TTP delay and OEF elevations. Furthermore, all patients with substantial TTP delays (≥4 seconds) had elevated OEF.19
A good agreement with the OEF has been shown for the ratio cerebral blood flow/cerebral blood volume in clinical studies as well as in experimental ischemia.17–19 Being the reciprocal of this ratio, MTT should also be an adequate predictor of OEF, because prolonged MTT as a consequence of hemodynamic impairment unites both a reduced cerebral perfusion and an autoregulatory vasodilatation with an increased cerebral blood volume. T2′ values showed a continuous decrease with the escalating MTT delay, which suggests that areas with different MTT delays reflect different degrees of OEF increase (Figure 2).
T2′ mapping is sensitive to artifacts caused by subject motion, especially in peripheral brain regions, for example, frontal or occipitoparietal regions.7,10 Apart from intrascan movement, artifacts can also be evoked by cerebrospinal fluid, both due to pulsation and partial volume effects. Several techniques were performed to minimize artifacts. First, we used a sophisticated motion correction algorithm as described previously by our group.10 In short, this algorithm is based on the principle that motion reduces the correlation between acquired and exponentially fitted data. The T2* data used for the calculation of T2′ images is acquired once at full resolution and twice at reduced resolution. Averaging these data sets with individually chosen weighting factors for each k-space line, the algorithm replaces k-space lines affected by intrascan movement and systematically reduces the influence of motion.10 Second, T2′ maps were thresholded excluding unphysiologically high (>300 ms) values and 0 voxels (<1 ms).13 Third, ROIs were placed in 3 consecutive slices of the periventricular white matter avoiding the inclusion of larger areas with cortical gray matter (risk of partial volume effects) or deep gray matter (basal ganglia, risk of T2′ differences due to physiological deposition of ferromagnetic substances). Fourth, we performed extensive control measurements to exclude unspecific, nonperfusion related T2′ differences by comparing ipsilateral and contralateral T2′ values in normoperfused MCA regions.
Our study has to deal with several limitations. Although our assumption that decreased T2′ values reflect increased oxygen extraction in perfusion-restricted areas is pathophysiologically plausible, a validation with the current imaging gold standard of oxygen metabolism, positron emission tomography will be necessary. Meanwhile, we cannot exclude that other factors than OEF changes contribute to differences in T2′ values, that is, variations in cerebral blood volume. This head-to-head comparison will also be useful for calibration of T2′ values to defined OEF levels. The interrelation of T2′ values and other parameters of perfusion disturbance (cerebral blood volume, cerebral blood flow, time to maximum) remains to be clarified. Additionally, our results are single-center results and an independent validation by others is of importance to assess feasibility and reproducibility of our MRI protocol. Although differences in T2′ values can occasionally be detected by visual inspection, due to the poor signal-to-noise ratio of T2′ maps, accurate results can only be achieved using ROI analysis. T2′ analysis of cortical gray matter is limited by partial volume effects. Although patients with asymptomatic and symptomatic stenoses differ substantially in terms of (recurrent) stroke risk, both were included in the study, because we focused on feasibility and technical issues of T2′ mapping with respect to perfusion abnormalities. Future assessment of the usefulness of T2′ mapping for stroke prediction and risk stratification will need larger patient samples with well-defined risk profiles.
In summary, motion-corrected T2′ mapping presumably detects areas with increased oxygen extraction within perfusion-disturbed tissue in patients with high-grade occlusive vessel disease. T2′ imaging may add important information to conventional MRI on the severity of the perfusion disturbance with respect to its metabolic consequences. A thorough comparison with the current gold standard for the assessment of cerebral oxygen metabolism, positron emission tomography is necessary before its general suitability and its value for clinical decision-making can be determined.
The online-only Data Supplement is available with this article at http://stroke.ahajournals.org/lookup/suppl/doi:10.1161/STROKEAHA.111.646109/-/DC1.
- Received November 23, 2011.
- Accepted March 29, 2012.
- © 2012 American Heart Association, Inc.
- Gorelick PB,
- Wong KS,
- Bae HJ,
- Pandey DK
- Derdeyn CP,
- Grubb RL Jr.,
- Powers WJ
- Baron JC,
- Bousser MG,
- Rey A,
- Guillard A,
- Comar D,
- Castaigne P
- Derdeyn CP,
- Videen TO,
- Yundt KD,
- Fritsch SM,
- Carpenter DA,
- Grubb RL,
- et al
- Ogawa S,
- Lee TM,
- Kay AR,
- Tank DW
- Geisler BS,
- Brandhoff F,
- Fiehler J,
- Saager C,
- Speck O,
- Rother J,
- et al
- Siemonsen S,
- Finsterbusch J,
- Matschke J,
- Lorenzen A,
- Ding XQ,
- Fiehler J
- Holst B,
- Siemonsen S,
- Finsterbusch J,
- Bester M,
- Schippling S,
- Martin R,
- et al
- Kluytmans M,
- van der Grond J,
- van Everdingen KJ,
- Klijn CJ,
- Kappelle LJ,
- Viergever MA
- Neumann-Haefelin T,
- Wittsack HJ,
- Fink GR,
- Wenserski F,
- Li TQ,
- Seitz RJ,
- et al
- Schumann P,
- Touzani O,
- Young AR,
- Morello R,
- Baron JC,
- MacKenzie ET
- Zaro-Weber O,
- Moeller-Hartmann W,
- Heiss WD,
- Sobesky J
- Kajimoto K,
- Moriwaki H,
- Yamada N,
- Hayashida K,
- Kobayashi J,
- Miyashita K,
- et al