Do Acute Diffusion- and Perfusion-Weighted MRI Lesions Identify Final Infarct Volume in Ischemic Stroke?
Background and Purpose— An acute mismatch on diffusion-weighted MRI (DWI) and perfusion-weighted MRI (PWI) may represent the “tissue-at-risk.” It is unclear which “semiquantitative” perfusion parameter most closely identifies final infarct volume.
Methods— Acute stroke patients underwent DWI and PWI (dynamic-susceptibility contrast imaging) on admission (baseline), and T2-weighted imaging (T2WI) at 1 or 3 months after stroke. “Semiquantitative” mean transit time (MTTsq=first moment of concentration/time curve), cerebral blood volume (CBVsq=area under concentration/time curve), and cerebral blood flow (CBFsq=CBVsq/MTTsq) were calculated. DWI and PWI lesions were measured at baseline and final infarct volume on T2WI acquired ≥1 month after stroke. Baseline DWI, CBFsq, and MTTsq lesion volumes were compared with final T2WI lesion volume.
Results— Among 46 patients, baseline DWI and CBFsq lesions were not significantly different from final T2WI lesion volume, but baseline MTTsq lesions were significantly larger. The correlation with final T2WI lesion volume was strongest for DWI (Spearman rank correlation coefficient ρ=0.68), intermediate for CBFsq (ρ=0.55), and weakest for MTTsq (ρ=0.49) baseline lesion volumes. Neither DWI/CBFsq nor DWI/MTTsq mismatch predicted lesion growth; lesion growth was equally common in those with and without mismatch.
Conclusions— Of the 2 PWI parameters, CBFsq lesions most closely identifies, and MTTsq overestimates, final T2WI lesion volume. “DWI/PWI mismatch” does not identify lesion growth. Patients without “DWI/PWI mismatch” are equally likely to have lesion growth as those with mismatch and should not be excluded from acute stroke treatment.
- cerebrovascular disorders
- imaging, diffusion-weighted
- imaging, perfusion-weighted
- magnetic resonance imaging
In acute ischemic stroke, the mismatch between the acute lesion on magnetic resonance (MR) diffusion-weighted imaging (DWI) and perfusion-weighted imaging (PWI) may represent the “ischemic penumbra” or “tissue-at-risk” of infarction1 and may provide a target of salvageable tissue within and possibly beyond 6 hours.2 In several studies, the presence of an acute PWI lesion larger than the DWI lesion at baseline predicted ischemic lesion growth on final (≥1 month) T2-weighted imaging (T2WI).3,4 One month is the minimum time at which final infarct extent should be assessed as “fogging” may obscure,5 and edema may exaggerate6 T2WI infarct extent measured at <1 month.
Perfusion information from MR dynamic-susceptibility contrast (DSC) PWI uses “semiquantitative” or “quantitative” analyses.7 We used “semiquantitative” (also called “summary” or “relative”7) to describe analyses of the concentration/time curve without correction with the arterial input function (AIF). In general, “semiquantitative” measures require less intensive processing than “quantitative” measures8 and include parameters such as: time-to-peak (TTP), which may be a good estimate of mean transit time (MTT)9 but is difficult to compare between imaging episodes;7 cerebral blood volume (CBVsq), calculated from the area under, and MTTsq, calculated from the first moment of, the concentration/time curve;7 and cerebral blood flow (CBFsq), calculated as CBVsq/MTTsq.10 “Quantitative” perfusion (also called “absolute”7) requires deconvolution of the concentration/time curve with an AIF:7,8 CBFq may be calculated as the height, and CBVq as the area under, the deconvolved concentration/time curve; and MTTq as CBVq/CBFq.7,8 CBV is not considered a good predictor of “tissue-at-risk” because of its bimodal nature and insensitivity to bolus delay and dispersion.11 Thus, parameters reflecting CBF and MTT have emerged as the most likely candidates for predicting final infarct extent.
Previous studies of PWI data (<24 hours after stroke) and final lesion extent (Table 1) have produced rather confusing results. Of the 13 studies, only 3 compared CBFq and MTTq (total n=53):11–13 2 found that CBFq and 1 that MTTq related best to final infarct volume. None directly compared CBFsq and MTTsq. In the remainder, TTP correlated strongly with final infarct volume4,6 and MTTsq and MTTq overestimated final infarct volume.14–21 Only 4 studies included patients who did not receive thrombolysis or experimental agents.11,12,15,20 The small sample sizes (Table 1), different PWI lesion measurement methods, inappropriate statistical tests (parametric versus nonparametric), and total absence of data comparing CBFsq and MTTsq (more rapidly calculated than “quantitative” data acutely) indicate that more data are required to determine which PWI parameter identifies “tissue-at-risk.”
We aimed to determine whether CBFsq or MTTsq on acute DSC PWI most closely identified the final T2WI infarct extent, the relationship of “DWI/PWI mismatch” and lesion growth, and to identify possible reasons for discrepancies in previous studies.
We prospectively recruited patients from the hospital stroke service with moderate to severe ischemic stroke. We performed imaging urgently (maximum for the study of 24 hours after stroke onset [baseline]). Baseline clinical assessment including the National Institutes of Health Stroke Scale (NIHSS) score and Oxfordshire Community Stroke Project (OCSP) classification were performed by a trained stroke physician. Functional outcome (modified Rankin scale [mRS]) was measured at 3 months. A panel of experts determined a final diagnosis of stroke. We recorded any acute treatments that the patients received (treatment did not affect inclusion in study).
We used a GE Signa LX 1.5T (General Electric) MR scanner with a self-shielding gradient set (22 mT/m maximum) and “birdcage” quadrature head coil. We included fast spin-echo T2WI and gradient-echo T1-weighted sequences, a spin-echo echo-planar (EP) imaging diffusion tensor protocol (DT-MRI), and DSC PWI. For DT-MRI, sets of axial diffusion-weighted EP images (b=0 and 1000 s/mm2) were collected with diffusion gradients applied sequentially along 6 noncollinear directions (5 acquisitions including baseline T2WI EP image and 6 diffusion-weighted EP images per slice position). PWI was measured by tracking gadolinium diethylenetriamine pentaacetic acid, injected intravenously by MR-compatible pump injector, for 85 s using a single-shot gradient-echo EP sequence. The DWI/PWI sequences used: 15 axial slices, 5-mm slice thickness, 1-mm slice gap, 128×128 image matrix, and 24×24 cm field-of-view.
“Semiquantitative” PWI parameters were calculated in each voxel from the signal intensities in the component EP images8 by fitting the data to a γ-variate curve. Maps of CBVsq (area under fitted concentration/time curve), MTTsq (first moment of fitted concentration/time curve), and CBFsq (CBVsq/MTTsq),8 were generated for each voxel, converted into Analyze (Mayo Foundation) format and presented as a color-coded map. Voxels in which the data did not reach a maximum value of >2× the maximum precontrast baseline signal SD and then begin to fall before the end of the imaging time (85 seconds) were assigned as error voxels, visible as white areas on the perfusion color maps (there were extremely few patients with any of these), indicating very low flow areas in the infarct.
Baseline PWI and DWI and Final T2WI Lesion Volume Analysis
A neuroradiologist, blind to all clinical and other imaging data, guided the tracing by an experienced neuroscientist of the visible lesions on DWI, CBFsq, and MTTsq. maps, using a Sun Ultra Sparc Station 10 (Sun Microsystems) in Analyze (ie, a consensus of 2 experienced observers). Image brightness and contrast were optimized between areas of abnormal diffusion or perfusion and normal-appearing brain. The neuroradiologist also traced the final infarct on the latter of the 1- or 3-month T2WI as described above. Lesion volumes were obtained by summing the number of outlined voxels and multiplying by the slice thickness on each slice on which the lesion was visible.
Baseline DWI, CBFsq, MTTsq, and final T2WI lesion volumes were not normally distributed (Kolmogorov–Smirnov test; P<0.01), so median lesion volumes were compared using Wilcoxon signed rank sum tests, and correlations between baseline DWI, CBFsq, and MTTsq volumes with final T2WI lesion volume were assessed with Spearman rank correlation coefficients (ρ). We also calculated mean volumes (±SD) for comparison with previous studies (Table 1), and Pearson product moment correlation coefficients (r, appropriate for parametric data) to compare baseline DWI, CBFsq, and MTTsq with final T2WI lesion volume, to determine whether use of Pearson r might explain differences between previous studies. “Lesion growth” was defined as final T2WI lesion volume>baseline DWI lesion volume, and “DWI/PWI mismatch” as PWI lesion volumes>DWI lesion volumes. χ2 tests were used to test whether DWI/CBFsq or DWI/MTTsq“mismatch” predicted “lesion growth.”
A total of 46 patients were included; 28 of 46 (61%) were male, with an average age of 72 years (range 37 to 94 years). Mean NIHSS score on admission was 8 (median 8; range 0 to 25). For OCSP classifications, 12 of 46 (26%) patients had a total anterior circulation infarct, 29 of 46 (63%) had a partial anterior circulation infarct, 4 of 46 (9%) had a lacunar infarct, and 1 of 46 (2%) had a posterior circulation infarct. All patients were treated with aspirin, but none received thrombolysis or investigational drugs. Mean mRS score at 3 months was 2 (median 2); 2 of 46 (4%) patients were dead (mRS=6), and 13 of 46 (28%) patients were dependent (mRS=3 to 5). Mean (±SD) time from stroke onset to baseline MRI was 10±7 hours; 18 of 46 (39%) patients were imaged within 6 hours, 11 of 46 (24%) at 6 to 12 hours, and 17 of 46 (37%) at 12 to 24 hours. Final T2WI was performed at 1 month (mean 32±4 days) in 6 of 46 (13%) patients and at 3 months (mean 96±7 days) in 40 of 46 (87%) patients.
Baseline PWI Lesion Volumes (Table 2)
Baseline CBFsq lesion volumes (median 23.74 cm3) were significantly smaller than baseline MTTsq lesion volumes (median 50.60 cm3; P<0.01). Eleven patients (24%) had no visible lesion on baseline CBFsq, and 10 patients (22%) had no visible lesion on baseline MTTsq; 6 (13%) had no visible lesion on either baseline CBFsq or MTTsq.
All patients had a visible baseline DWI lesion. Baseline DWI lesion volumes (median 13.28 cm3) were significantly smaller than baseline CBFsq (P=0.01) and MTTsq volumes (P<0.01) but were not significantly different from final T2WI lesion volume (P=0.14).
All patients had a visible lesion on final T2WI (median 18.47 cm3) corresponding with the lesion indicated on baseline DWI. Overall, baseline CBFsq lesion volume was not significantly different from final T2WI lesion volume (P=0.82). Baseline MTTsq volume was significantly larger (P<0.01) than final T2WI lesion volume.
Final T2WI lesion volume most strongly correlated with baseline DWI (ρ=0.68), followed by baseline CBFsq (ρ=0.55) and baseline MTTsq (ρ=0.49) lesion volumes (all P<0.01). Using Pearson’s r, final T2WI lesion volume was most strongly correlated with baseline MTTsq (r=0.66), followed by baseline DWI (r=0.53) and baseline CBFsq (r=0.41) lesion volumes (all P<0.01).
“DWI/PWI Mismatch” and Final T2WI Lesion Volume
There was no significant association between “lesion growth” (final T2WI lesion>baseline DWI lesion) and the presence or absence of “DWI/PWI mismatch” (Table 3). On CBFsq, 15 of 25 patients with “mismatch” had “lesion growth,” and 10 of 25 patients did not; 11 of 21 without “mismatch” had “lesion growth,” and 10 of 21 did not (χ2=0.27; P=0.60; Table 3). On MTTsq, 21 of 33 with “mismatch” had “lesion growth,” and 12 of 33 did not; 5 of 13 without “mismatch” had “lesion growth,” and 8 of 13 did not (χ2=2.40; P=0.12; Table 3). Thus, about half of patients without “mismatch” had lesions that grew, and “mismatch” was as common among patients with lesions that grew as among those with lesions that did not grow (Figure 2).
Because there were no previous studies testing the relationship between acute CBFsq and MTTsq and final infarct volume, we asked: which “semiquantitative” PWI parameter, if any, best predicts final infarct extent? In this large systematic study, we found that acute CBFsq lesions more closely identified final T2WI lesion extent than acute MTTsq lesions, which clearly overestimate final lesion extent. We also demonstrate that “lesion growth” was equally common among patients with and without “DWI/PWI mismatch”: half the patients without a “mismatch” had “lesion growth”; half of those with “mismatch” did not. Although “DWI/PWI mismatch” is being used to identify patients for inclusion in trials of stroke treatments, our results imply that patients without “mismatch” may also benefit and should not be denied that possibility by being excluded.
Our findings concerning “semiquantitative” PWI parameters agree overall with the few previous “quantitative” studies directly comparing CBFq and MTTq.11–13 Some disparities among previous studies may result from different statistical analyses; use of Pearson r instead of Spearman ρ completely changed the strength and order of correlation between baseline CBFsq and MTTsq lesions and final T2WI lesion volume. Previous studies also support the conclusion that acute MTT (MTTsq or MTTq) lesions overestimate final lesion extent,12–21 although some studies found stronger correlations between MTT and final infarct volumes than we did (eg, ρ=0.77 to 0.9012,13,19,20); these studies used MTTq, were small (mean n=21; Table 1), and tested some very restricted perfusion thresholds (eg, highest 70% of MTT signal intensities12). Furthermore, it is important to distinguish between “correlation” and “size equivalence”; a good correlation between baseline MTT volume and final T2WI lesion volume is not incompatible with MTT significantly overestimating final infarct volume.
In contrast to some previous studies,4,11–13,16,19,20 we found no significant association between DWI/CBFsq or DWI/MTTsq “mismatch” and “lesion growth” (Table 3). However, some of these studies only included patients with “mismatch.”11,12 A recent study using “DWI/PWI mismatch” as an inclusion criteria in a thrombolysis trial22 failed to show an association between “DWI/PWI mismatch” and lesion expansion but did show that reperfusion was inversely associated with “lesion growth.”
The superiority of “quantitative” over “semiquantitative” PWI measures is not yet proven. “Quantitative” measures are still only estimates7 and require considerably more complex image processing than “semiquantitative” measures so are less appropriate to the acute setting. Newer perfusion measures including flow heterogeneity21 and bolus-delay corrected maps13 appear promising, but further validation is required to determine their overall usefulness. There is a further question concerning the use of the term “perfusion-weighted imaging” itself, which is somewhat misleading because MR perfusion maps are actually calculated parameter maps.
The strengths of the current study include the large sample, inclusion of patients with and without “DWI/PWI mismatch,” careful clinical characterization, blinded image analysis and outcome assessment, and careful use of statistics. We included patients up to 24 hours, and interestingly most (75% to 80%) had a perfusion lesion on CBF, MTT, or both, with only just >10% not having a perfusion lesion. We used manual rather than “threshold” measurement methods because the former is currently the most rapid method to apply in the clinical setting. Thresholds might reduce observer variation, but the hemodynamics underlying ischemic stroke are complex and variable. Baseline perfusion values greatly overlap between tissue that does and does not infarct,12,17 and PWI parameter thresholds are time dependent; a threshold for ischemic but viable brain at 3 hours may be very different from that at 6 hours. None of our patients received thrombolysis or any investigational drugs, therefore this series represents a baseline measure of the presence or absence of spontaneous lesion expansion and could be used to estimate sample sizes for future acute stroke treatment trials using imaging markers of outcome. However, our data emphasize the need for further study to identify which, if any, alternative imaging markers might identify patients in whom treatments such as thrombolysis might prevent infarct growth. For example, it is unclear precisely how large artery patency relates to the PWI lesion. Are middle cerebral artery or internal carotid artery occlusions always accompanied by PWI lesions, and do these lesions resolve completely and rapidly as the artery reperfuses, or is there a time lag between arterial reperfusion and PWI lesion disappearance?
In conclusion, acute DWI, CBFsq, and MTTsq lesion volumes all correlated well with final infarct volume, but acute MTTsq lesion volume significantly overestimated final infarct volume. Of the 2 “semiquantitative” PWI measures, the acute CBFsq lesion most closely identifies the final T2WI lesion. There was no clear association between the presence of a “DWI/PWI mismatch” and lesion expansion, questioning the value of the “DWI/PWI mismatch” as a clinical marker of “tissue-at-risk.” About half of patients without “DWI/PWI mismatch” have “lesion growth” so they may benefit from acute stroke treatments and should probably not be excluded from clinical trials.
This study was funded by the Scottish Executive’s Chief Scientist Office (Reference No. C2B/4/14) and the Row Fogo Charitable Trust. C.S.R. is funded by a Royal Society of Edinburgh/Lloyds TSB Foundation studentship. The work was conducted at the SHEFC Brain Imaging Research Centre for Scotland.
- Received June 27, 2005.
- Revision received September 22, 2005.
- Accepted October 13, 2005.
Schlaug G, Benfield A, Baird AE, Siewert B, Lovblad KO, Parker RA, Edelman RR, Warach S. The ischemic penumbra: operationally defined by diffusion and perfusion MRI. Neurology. 1999; 53: 1528–1537.
Albers GW. Expanding the window for thrombolytic therapy in acute stroke. The potential role of acute MRI for patient selection. Stroke. 1999; 30: 2230–2237.
Barber PA, Darby DG, Desmond PM, Yang Q, Gerraty RP, Jolley D. Prediction of stroke outcome with echoplanar perfusion- and diffusion-weighted MRI. Neurology. 1998; 51: 418–426.
Teng MM, Cheng HC, Kao YH, Hsu LC, Yeh TC, Hung CS, Wong WJ, Hu HH, Chiang JH, Chang CY. MR perfusion studies of brain for patients with unilateral carotid stenosis or occlusion: evaluation of maps of “time to peak” and “percentage of baseline at peak.” J Comput Assist Tomogr. 2001; 25: 121–125.
Zierler KL. Equations for measuring blood flow by external monitoring of radioisotopes. Circ Res. 1965; 16: 309–321.
Rohl L, Ostergaard L, Simonsen CZ, Vestergaard-Poulsen P, Andersen G, Sakoh M, Le Bihan D, Gyldensted C. Viability thresholds of ischemic penumbra of hyperacute stroke defined by perfusion-weighted MRI and apparent diffusion coefficient. Stroke. 2001; 32: 1140–1146.
Parsons MW, Yang Q, Barber PA, Darby DG, Desmond PM, Gerraty RP, Tress BM, Davis SM. Perfusion magnetic resonance imaging maps in hyperacute stroke: relative cerebral blood flow most accurately identifies tissue destined to infarct. Stroke. 2001; 32: 1581–1587.
Rose SE, Janke AL, Griffin M, Finnigan S, Chalk JB. Improved prediction of final infarct volume using bolus delay-corrected perfusion-weighted MRI: implications for the ischemic penumbra. Stroke. 2004; 35: 2466–2471.
Ueda T, Yuh WT, Maley JE, Quets JP, Hahn PY, Magnotta VA. Outcome of acute ischemic lesions evaluated by diffusion and perfusion MR imaging. Am J Neuroradiol. 1999; 20: 983–989.
Butcher K, Parsons M, Baird T, Barber A, Donnan G, Desmond P, Tress B, Davis S. Perfusion thresholds in acute stroke thrombolysis. Stroke. 2003; 34: 2159–2164.
Rohl L, Geday J, Ostergaard L, Simonsen CZ, Vestergaard-Poulsen P, Andersen G, Le Bihan D, Gyldensted C. Correlation between diffusion- and perfusion-weighted MRI and neurological deficit measured by the Scandinavian Stroke Scale and Barthel Index in hyperacute subcortical stroke (≤6 hours). Cerebrovasc Dis. 2001; 12: 203–213.
Rose SE, Chalk JB, Griffin MP, Janke AL, Chen F, McLachan GJ, Peel D, Zelaya FO, Markus HS, Jones DK, Simmons A, O’Sullivan M, Jarosz JM, Strugnell W, Doddrell DM, Semple J. MRI based diffusion and perfusion predictive model to estimate stroke evolution. Magn Reson Imaging. 2001; 19: 1043–1053.
Hacke W, Albers G, Al Rawi Y, Bogousslavsky J, Davalos A, Eliasziw M, Fischer M, Furlan A, Kaste M, Lees KR, Soehngen M, Warach S. The Desmoteplase in Acute Ischemic Stroke Trial (DIAS): a phase II MRI-based 9-hour window acute stroke thrombolysis trial with intravenous desmoteplase. Stroke. 2005; 36: 66–73.