Improved Prediction of Final Infarct Volume Using Bolus Delay–Corrected Perfusion-Weighted MRI
Implications for the Ischemic Penumbra
Background and Purpose— Magnetic resonance imaging (MRI)-based perfusion measures using dynamic susceptibility contrast are extremely useful for identification of ischemic penumbral tissue in acute stroke. However, errors in the measurement of cerebral blood flow (CBF) and mean transit time (MTT) can occur. The aim of this study was to investigate whether bolus delay–corrected (BDC) perfusion measures enable better delineation of the ischemic penumbra.
Methods— Diffusion-weighted MRI (DWI) and perfusion-weighted MRI data were acquired from 19 acute stroke patients. Perfusion abnormalities were manually defined on BDC perfusion maps (corrected MTT [cMTT] and corrected CBF [cCBF]), and on maps derived from an arterial input function placed within the contralateral (CBF, MTT) and ipsilateral (ipsilateral CBF [iCBF] and ipsilateral MTT [iMTT]) middle cerebral artery. Perfusion lesion volumes were correlated with 30-day T2-weighted MRI lesion volumes and with clinical outcome using the National Institutes of Health Stroke Scale (NIHSS).
Results— Spearman correlation coefficients for comparing lesion volumes delineated on DWI, CBF, iCBF, cCBF, MTT, iMTT, and cMTT maps with 30-day T2-weighted lesion volumes were 0.72, 0.87, 0.88, 0.90, 0.84, 0.92, and 0.96, respectively (all P<0.001). The analogous correlation coefficients for comparing 30-day National Institutes of Health Stroke Scale (NIHSS) scores were 0.39 (NS), 0.69 (NS), 0.75 (P<0.001), 0.62 (NS), 0.72 (P<0.001), 0.78 (P<0.001), and 0.83 (P<0.001), respectively.
Conclusions— Uncorrected perfusion lesion volumes overestimated the extent of ischemic injury. BDC perfusion measures (cMTT) correlated more accurately with final lesion volume and clinical outcome. Such measures offer an improved estimation of the final infarct size in acute stroke.
- cerebral ischemia
- magnetic resonance imaging, diffusion-weighted
- magnetic resonance imaging, perfusion-weighted
- stroke, acute
Accurate measurement of the ischemic penumbra is important for clinical management of acute stroke patients, for evaluation of the efficacy of new thrombolytic or neuroprotective agents, and to improve our understanding of the pathophysiology associated with infarct evolution. Recently, Kidwell1 described a new model of the ischemic penumbra that includes both regions of the diffusion–perfusion mismatch as well as portions of the diffusion lesion. Estimation of the size of this region is dependent on accurate estimation of MRI perfusion measures.
Using MRI techniques, cerebral perfusion is normally measured using dynamic susceptibility contrast (DSC).2 Because of the established coupling between contrast mean transit time (MTT) and cerebral perfusion pressure,3 the perfusion abnormality is commonly assessed using the MTT or with surrogate markers describing the bolus contrast hemodynamic function, such as time-to-peak4 or Tmax.1 Recent theoretical analysis of the DSC technique has shown that significant error in the quantitation of cerebral blood flow (CBF) and MTT can occur in patients with cerebral vascular disease caused by delay and dispersion of the bolus of contrast agent.5,6⇓ The error in these perfusion measures can be reduced by using bolus delay–corrected (BDC) methods5,6⇓ or circular convolution techniques.7
The aim of this study was to investigate whether BDC MTT measures (cMTT) enabled better prediction of the region destined to undergo infarction. This was performed by comparing manually defined abnormal regions on cMTT and uncorrected MTT maps, obtained with arterial input functions (AIFs) placed within either the contralateral (MTT) or ipsilateral (iMTT) middle cerebral arteries (MCAs), with T2-weighted MRI lesion volumes measured at 30 days after stroke. In addition, these MTT indices were correlated with follow-up clinical outcome measures. Similar correlations were also performed on volumes of abnormal perfusion defined on CBF maps. We postulate that cMTT measures that more accurately reflect the cerebral hemodynamic status in the acute stroke setting will enable an improved estimation of the extent of the ischemic penumbra.
Materials and Methods
Nineteen patients (10 male, 71.3 ± 11.3 years of age) with acute focal neurological symptoms consistent with ischemic cortical stroke were recruited. Approval to conduct the study was obtained from the local university and hospital human experimental ethics committees. The patient or legally authorized person gave informed consent for the study. The “time of first scan” was defined as the time elapsed between the last time the patient was known to be without neurological deficit (accurate to within 1 hour) and the time of the initial MRI scan. Patients were excluded if they presented with a cerebral hemorrhage on computed tomography scan or some other pre-existing neurological condition that would confound clinical MR assessment, such as history of previous stroke. Four patients received thrombolytic therapy with intravenous recombinant tissue plasminogen activator (r-tPA). For these patients, MRI scans were performed before administration of r-tPA. Follow-up magnetic resonance angiogram (MRA) examination revealed no visible recanalization of the affected MCA in any case. The National Institutes of Health Stroke Scale (NIHSS) was obtained at all imaging time points.
MRI Data Acquisition
All patients received serial diffusion tensor (DTI), perfusion (DSC), T2, and MRA examinations using either a 1.5-T Siemens Sonata or a 1.5-T GE Echospeed MRI scanner. Similar pulse sequence acquisition parameters were used on both systems. The maximum gradient strengths for the scanners were 40 and 25 mT/m, respectively. An optimized DTI sequence8 was used with the following parameters: 21 axial slices; field of view (FOV) 23 cm; repetition time (TR) 4.2 seconds; echo time (TE) 106 milliseconds; 5-mm slice thickness with 1.5-mm gap and 30 b values per direction (7 gradient directions, 22 high [b=1096 seconds/mm] and 8 low b values [b=0]). Perfusion maps were obtained using dynamic fast bolus tracking of OptiMark (0.2 mL/kg; injection rate 5 mL/s, gadoversetamide; Mallinckrodt) using a spin echo echo-planar imaging sequence. The imaging parameters were 19 axial slices; FOV 23 cm; TR 2 seconds; TE 60 milliseconds; 5-mm slice thickness with 1.5-mm gap; with an acquisition of 50 frames per slice. Uncorrected MTT maps were calculated using the previously described method of Ostergaard.2 For these maps, the AIF was selected from the contralateral and ipsilateral MCA to the cerebral infarct. Pixels chosen to represent the AIF showed a large increase in intensity on the concentration time curve compared with normal brain parenchyma.9 Uncorrected perfusion images derived from the contralateral and ipsilateral MCA were labeled CBF, MTT, iCBF, and iMTT, respectively. BDC perfusion maps were generated from an AIF placed within the contralateral MCA. Volumes of the acute diffusion, perfusion, and 30-day T2-weighted MRI lesion volumes were determined by operator-defined manual tracing.
To correct for bolus delay, we used a technique whereby the bolus concentration time course was shifted to coincide with that of the AIF.5 This was achieved using a simple geometric-based algorithm. A schematic diagram outlining this procedure is given in Figure 1. For every pixel within the brain, a line was projected from the maximum of the bolus concentration time curve to the start of the acquisition of the time series. The bolus arrival time was obtained by determining the point on the initial rise of the concentration time curve furthest from the projected line. The bolus delay was defined as the change in the arrival time between the AIF and the voxel of interest. For each voxel, the time course was shifted by an integral value of the TR to closely match that of the AIF. After correction of the bolus delay, perfusion measures were evaluated using standard deconvolution methods.2
For delineation of lesion volumes, images were assessed twice with at least a 2-week period between assessments by a single rater (stroke neurologist) with experience in manually tracing diffusion and perfusion lesion volumes. The rater was blinded to all clinical information and had no knowledge of whether perfusion maps were uncorrected or corrected for bolus delay. Lesions were traced using the DISPLAY image processing software (Montreal Neurological Institute) and windowing was adjusted by the rater. Reproducibility of the measure was assessed with ANOVA and the Bland–Altman test for comparing different measures of the same quantity (ie, to test whether the mean of the measures is within 2 SDs of the mean difference). Spearman correlation coefficients were computed to evaluate the correlation between the initial diffusion-weighted MRI (DWI), CBF, iCBF, cCBF, MTT, iMTT, and cMMT lesion volumes with 30-day T2-weighted MRI lesion volume and outcome NIHSS score measured at 30 days. Bonferroni correction for multiple comparisons was applied to maintain the total Type I error rate at a sufficiently low level. In this case, 10 comparisons were performed. A P<0.003 was considered statistically significant.
Patient demographics, time of initial scan, and clinical outcomes scores are given in Table 1. All patients in this study had ischemic lesions in the MCA territory. The mean initial scan time was 4.3±1.6 hours. There was no significant difference between lesion volumes when assessing intrarater reliability as measured by ANOVA or the Bland–Altman test for reproducibility of measure. Pearson correlation coefficients for the diffusion and perfusion lesion volumes for each intrarater measure were >0.98. Acute diffusion and perfusion lesion size along with follow-up 30-day T2-weighted volume for the patient cohort are listed in Table 2. A summary of the results of the correlation between acute diffusion and perfusion lesion volumes with follow-up 30-day T2-weighted lesion volume and clinical outcomes scores are given in Table 3.
Shown in Figure 2 (patient 1) are representative perfusion maps windowed at the same threshold level demonstrating the effects of bolus delay correction. There is increased blood flow in normal parenchymal tissue and within the penumbral region on BDC maps (cCBF) compared with corresponding uncorrected maps (cCBF, iCBF). In similar regions, BDC cMTT maps reveal reduced MTT compared with uncorrected maps (MTT, iMTT). Such perfusion changes are in agreement with theoretical analyses reported previously5 and clinical results regarding BDC strategies.6,7⇓ Analysis of the volumetric data revealed that regions of abnormal perfusion on uncorrected MTT and iMTT maps significantly overestimated the volume of the eventual infarct size in the acute stroke setting. As demonstrated in Figure 2, uncorrected CBF maps generated from an AIF placed within either the contralateral (CBF) or ipsilateral (iCBF) MCA show an area of possible hypoperfusion extending from the MCA watershed territory into the posterior unaffected hemisphere. The uncorrected MTT and iMTT maps also reveal a significant perfusion abnormality in the posterior unaffected hemisphere. In contrast, BDC perfusion maps (cCBF, cMTT) revealed a much smaller volume of tissue with abnormal hemodynamic function that better reflected the extent of neuronal injury shown on the 30-day T2-weighted MRI scan. Of the 18 patients with an initial diffusion lesion >10 mL, 14 patients possessed 30-day T2-weighted infarct volumes within 10% of the volume of the perfusion abnormality delineated on acute cMTT maps. Perfusion images for a representative patient 2 with large acute cMTT perfusion lesion/final infarct lesion volume mismatch are given in Figure 3. Such a mismatch may represent a region of benign oligemia. Although MRI scans from 4 patients (2, 5, 15, and 16) revealed the presence of such regions, perfusion lesion volumes on cMTT maps more significantly correlated with the NIHSS scores than did uncorrected MTT or any CBF measure.
When comparing uncorrected perfusion measures, there was a stronger correlation with lesion volume measured from iCBF and iMTT maps than with CBF or MTT maps when compared with the 30-day T2-weighted MRI infarct size. However, in 1 patient (16), the iCBF and iMTT maps revealed regions of tissue with abnormal perfusion within the unaffected hemisphere.
In the present study, a single expert observer, who was blinded to all clinical and image information, measured stroke lesion volumes. We found a high degree of reproducibility between measures, consistent with the results of previous studies reporting that operator-defined measurements of the diffusion/perfusion mismatch are reproducible and useful for assessing patient clinical status and determining therapeutic efficacy.9–11⇓⇓
The most significant finding of this study was that the volume of tissue with abnormal hemodynamic function on the cMTT map was more strongly correlated with the infarct volume on the 30-day T2-weighted MRI scans than was the MTT perfusion lesion. Lesion volumes on the cMTT maps were also correlated more highly with the 30-day NIHSS scores than were lesion volumes derived from MTT maps. Perfusion lesion volumes measured using MTT maps significantly overestimated the extent of the final infarct size in the acute stroke setting. Although this finding has been expected from the theoretical analysis of the DSC method,5–7⇓⇓ this is the first study to demonstrate the importance of using BDC perfusion measures in acute stroke patients. Although the accuracy of DSC-based perfusion measures can be improved using various correction strategies,2,5,7⇓⇓ we used a computationally simple geometric approach whereby the bolus concentration time course for each pixel was shifted to coincide with that of the AIF.5 Further improvement in perfusion measures can be achieved by correcting the dynamic bolus time course function for the effects of tracer dispersion within a voxel.12
With respect to uncorrected perfusion measures, blood flow and MTT maps generated from an AIF placed within the ipsilateral MCA (iCBF, iMTT) better correlated with follow-up infarct size and clinical outcome compared with uncorrected CBF and MTT measures derived from AIF selection within the contralateral MCA. This finding is in agreement with previous reported studies,6,10⇓ with the reduction in error resulting from the inherent delay of the contrast agent within the ipsilateral MCA. In a recent report, Thijs et al11 investigated the relationship between hemodynamic lesion volume, follow-up infarct size, and location of AIF placement in a cohort of acute stroke patients. They determined that perfusion lesion volume derived from an AIF placed within the contralateral MCA better correlated with follow-up infarct size. However, it should be noted that this study used a different measure of perfusion (Tmax) and that the final infarct size was determined using a 4- to 6-day DWI lesion volume rather than a T2-weighted MRI scan at 30 days after stroke.
Although we advocate using ipsilateral MCA-generated CBF and MTT measures, if bolus correction methods are not available, care must be taken when interpreting hemodynamic function. One patient (16) possessed a perfusion abnormality that extended into the unaffected hemisphere. Such a result can occur when selecting an AIF within the MCA in close proximity to the infarct because the bolus concentration time curve for a given pixel can precede in time the AIF.6 In a similar fashion, stenotic disease of the internal carotid artery can also affect bolus arrival times and yield erroneous perfusion measures.13 Use of BDC techniques reduces the ambiguity regarding appropriate selection of the AIF.
New models describing the penumbra now include regions of hypoperfused neuronal tissue that do not progress to infarction, namely regions of benign oligemia.1 Improved delineation of the perfusion lesion by BDC permits investigation of such regions. In our cohort of stroke patients, 14 of the 18 patients with an initial diffusion lesion >10 mL possessed final infarct volumes nearly equal to the volume of the perfusion abnormality delineated on acute cMTT maps. Four patients (2, 5, 15, and 16) possessed abnormal perfusion volumes on acute maps that were significantly larger than 30-day T2 lesion volumes. Of these patients, 15 and 16 received thrombolytic therapy with r-tPA (intravenously). Although no vessel recanalization was observed during follow-up MRA examination, some beneficial effect of thrombolysis may have occurred that was below the limits of resolution of the MRA examination. Patient 5 had clinical evidence of spontaneous reperfusion in the period between hospital admission and initial MRI examination (NIHSS scores reduced from 7 to 2 during the first 6 hours). Patient 2 did not receive any thrombolytic or neuroprotective therapy, although his NIHSS score also reduced from an initial score of 25 to 15 during a 16-hour period. Perfusion maps, identically windowed at 2 different threshold levels to aid delineation of the lesion (0 to 20 seconds [cMTT, MTT] and 3 to 16 seconds [cMTT*, MTT*]) are shown in Figure 3. It is apparent that the abnormal perfusion seen on the uncorrected maps is considerably larger than the final infarct size and overestimates acute neuronal injury. The BDC perfusion map windowed at the lower threshold level (cMTT*) also shows regions of tissue with abnormal hemodynamic function that are larger than the infarct volume represented on the 30-day T2-weighted MRI scan but at a much reduced volume. Although the central core of this perfusion lesion matches the final infarct, there exists an area of subtly hypoperfused tissue that does not progress to infarction. Some of this territory may represent benign oligemic tissue. In the present study, uncorrected perfusion measures overestimated the volume of hypoperfused tissue that has the potential to survive the ischemic event. Use of such uncorrected measures may bias evaluation of the efficacy of new drug therapies. The results of this study support the concept of a penumbra containing benign oligemic tissue in some patients; many patients in this study appeared to have a region of benign oligemia on uncorrected measures that were not present on corrected MRI scans. Further studies using BDC MRI involving larger patient cohorts are required to fully elucidate our understanding of the ischemic penumbra.
We acknowledge funding from GlaxoSmithKline Pharmaceuticals, United Kingdom (G.S.K.), and the support of the Stroke Units of the Royal Brisbane, Princess Alexandra, and The Wesley Hospitals, Australia.
- Received May 30, 2004.
- Revision received August 26, 2004.
- Accepted August 31, 2004.
Kidwell CS, Alger JR, Saver JL. Beyond mismatch: evolving paradigms in imaging the ischemic penumbra with multimodal magnetic resonance imaging. Stroke. 2003; 34: 2729–2735.
Sette G, Baron JC, Mazoyer B, Levasseur M, Pappata S, Crouzel C. Local brain hemodynamics and oxygen metabolism in cerebrovascular disease. Brain. 1989; 112: 931–951.
Neumann-Haefelin T, Wittsack HJ, Wenserski F, Siebler M, Seitz RJ, Modder U, Freund HJ. Diffusion and perfusion weighted MRI: the DWI/PWI mismatch region in acute stroke. Stroke. 1999; 30: 1591–1597.
Calamante F, Gadian DG, Connelly A. Quantification of perfusion using bolus tracking magnetic resonance imaging in stroke. Assumptions, limitations and potential implications for clinical use. Stroke. 2002; 33: 1146–1151.
Coutts SB, Simon JE, Tomanek AI, Barber PA, Chan J, Hudon ME, Mitchell JR, Frayne R, Eliasziw M, Buchan AM, Demchuk AM. Reliability of assessing percentage of diffusion-perfusion mismatch. Stroke. 2003; 34: 1681–1685.
Yamada K, Wu O, Gonzalez G, Bakker D, Ostergaard L, Copen WA, Weisskoff RM, Rosen BR, Yagi K, Nishimura T, Sorensen AG. Magnetic resonance perfusion-weighted imaging of acute cerebral infarction. Effect of the calculation methods and underlying vasculopathy. Stroke. 2002; 33: 87–94.
Thijs VN, Somford DM, Bammer R, Robberecht W, Moseley ME, Albers GW. Influence of arterial input function on hyperperfusion volumes measured with perfusion-weighted imaging. Stroke. 2004; 35: 94–98.