Abstract
Background: Identifying tissue at risk for infarction is important in deciding which patients would benefit most from potentially harmful therapies and provides a way to evaluate newer therapies with regard to the amount of ischemic tissue salvaged.
Objective: To operationally define and characterize cerebral tissue at risk for stroke progression.
Methods: We retrospectively selected 25 patients with an acute onset of a hemispheric stroke from our database who had undergone a combination of two diffusion-weighted MRI studies and a perfusion-weighted MRI study. We applied a logistic regression model using maps of the relative mean transit time and relative cerebral blood flow (rCBF) as well as three different maps of the relative cerebral blood volume (rCBV) to predict an operationally defined penumbra (region of mismatch between the diffusion lesion on day 1 and its extension 24 to 72 hours later).
Results: Maps of the rCBF and initial rCBV were significant predictors for identifying penumbral tissue. Our operationally defined penumbral region was characterized by a reduction in the initial rCBV (47% of contralateral control region [CCR]), an increase (163% of CCR) in the total rCBV, and a reduction (37% of CCR) in the rCBF, whereas the operationally defined ischemic core showed a more severe reduction in the rCBF (12% of CCR) and in the initial rCBV (19% of CCR).
Conclusion: These MR indexes may allow the identification and quantification of viable but ischemically threatened cerebral tissue amenable to therapeutic interventions in the hyperacute care of stroke patients.
The ischemic penumbra is a region of incomplete ischemia adjacent to the zone of complete ischemia, the core of the ischemic infarct. As originally defined, it encompasses that portion of the oligemic territory where electrical failure has occurred but cellular integrity is maintained.1 The concept of an ischemic penumbra has been a major impetus for developing pharmacologic interventions in the acute treatment of stroke patients to potentially salvage a peri-infarct region that is at high risk for infarct evolution. The identification of amenable tissue in the hyperacute period is of critical importance because the tolerance of perfusional disturbances is related to its duration, which can determine the progression of ischemia from the core into the oligemic penumbral region. It has been shown in experimental animal stroke models that the penumbra remains viable for some time and that the infarct core gradually expands into the ischemic penumbra.2-6 In human studies it was found that viable tissue in the ischemic penumbra can be found up to 48 hours after stroke onset,3,7,8 which would make the time window for therapeutic interventions considerably longer than the currently hypothesized 4 to 6 hours.
Factors that contribute to the deterioration of the penumbra and evolution of the ischemic core region are misery perfusion, excitotoxicity, acidosis, recurrent spreading depression, and others.5,9,10 It is still a matter of debate whether energy metabolism is preserved5 or transiently disturbed in the penumbral region.11 In the clinical setting, only a few of the many factors contributing to this complex process of ischemic cell damage can be assessed with PET or newer MRI methods such as diffusion and perfusion techniques.
PET studies have used combinations of low cerebral blood flow (CBF), relatively preserved or normal cerebral metabolic rate of oxygen (CMRO2), and a high oxygen extraction fraction (OEF) to distinguish tissue at risk for infarction but potentially recoverable from infarcted tissue that was characterized by a severely reduced CBF, low CMRO2, and a low OEF.3,5,7,8,12,13
Although the new MR techniques of diffusion and perfusion imaging enable us to rapidly identify ischemic and oligemic tissue,14-19 they have not been applied to identify ischemically threatened but not yet infarcted tissue. Diffusion as well as perfusion imaging both measure unique and complimentary events in the ischemic cascade, and combining both techniques could help differentiate between ischemic core and penumbra. A decline in the apparent diffusion coefficient (ADC), which is the quantitative expression of tissue diffusivity, is associated with a decrease of the energy requiring Na+-K+-ATPase20-22 and as such may permit the identification of the ischemic core, a region of energy failure. An enlargement of the ischemic core is frequently observed in the first 24 to 48 hours in human stroke. This enlargement occurs within the hypoperfused region, and in our model of the ischemic penumbra it is proposed that this enlargement represents the ischemic penumbra.
The intent of this study is to investigate 1) whether a combination of diffusion-weighted MRI (DWI) and perfusion-weighted MRI (PWI) allows an operational definition of cerebral tissue at risk for infarction, and 2) whether MR values of relative cerebral blood volume (rCBV), relative CBF (rCBF), and water diffusivity can be derived to characterize operationally defined ischemic core and penumbra regions. These characterizations could be important in identifying tissue amenable for thrombolytic or neuroprotective agents in the hyperacute stage of stroke evolution.
Methods.
Patients.
This study consists of 25 patients (15 men and 10 women, mean age 69 years) taken from our database of acute stroke patients. To fulfill the criteria for inclusion in this study, 1) patients had to present with an acute onset of neurologic signs and symptoms suggestive of a hemispheric ischemic stroke, 2) DWI and PWI studies had to be performed within 24 hours after onset, and 3) patients had to have follow-up DWI showing an increase in lesion size within 24 to 72 hours after the first MRI study. Patients were retrospectively selected from our database, covering a time period of 30 months. Seventy-two patients fulfilled inclusion criteria with regard to the time windows. Fourteen of these patients had to be excluded due to not showing an increase in the DWI abnormality. Eighteen patients had to be excluded because they had small vessel strokes and did not fulfill the criteria of having a hemispheric stroke. Fifteen patients were excluded because a perfusion study had not been performed or because of technical problems during the imaging.
The stroke onset time was taken from the time the patient was last known to be without the new deficits. If a patient woke up with a deficit, the time the patient went to bed or was last seen to be functioning normally was taken as stroke onset time. Only 3 of these 25 patients woke up with a deficit, and their stroke onset time may be off by approximately 8 hours. Patients may have been treated with anticoagulant or antiplatelet drugs, and 3 of our 25 patients were treated with recombinent tissue plasminogen activator, but inclusion or analysis based on treatment was beyond the scope of this study. All patients or their authorized representative gave written informed consent to the MRI procedure, which was approved by the Committee on Clinical Investigations of the Beth Israel Deaconess Medical Center.
MRI.
In each of the patients, DWI and PWI were performed. Some of the MRI studies (n = 6) were performed using a prototype whole-body, 1.5-tesla, echo-planar imaging (EPI)-capable system (Siemens Medical Systems, Erlangen, Germany). The remaining studies (n = 19) were performed on a Siemens Vision 1.5-tesla EPI system. A circularly polarized head coil was used for excitation and signal reception. In addition to diffusion and perfusion MR images, conventional spin-echo T2-weighted images (T2WIs), proton-density–weighted images, and T1-weighted images were acquired. All MRI studies were acquired within 15 to 25 minutes, with DWI always preceding PWI.
Diffusion-weighted MRI.
On the prototype whole-body 1.5-tesla EPI system, DWI was performed using a multislice, single-shot, spin-echo EPI sequence.15,18 The diffusion gradient was applied in the transverse (x) direction, with seven b values ranging from 0 to 1,271 sec/mm2. On the Siemens Vision 1.5-tesla EPI system, we used only two b values (0 and 1,000 sec/mm2). Typical imaging parameters were an echo time (TE) of 118 msec, matrix size of 128 × 128, field of view (FOV) of 260 mm, slice thickness of 7 mm (no gap between slices), and a set of 20 axial slices. The MR diffusion sequence at b = 1,000 was run three times, with diffusion gradients applied in each of the x, y, and z directions. To minimize the effects of diffusion anisotropy, an average of all three diffusion directions was calculated to give the trace of the diffusion tensor. The calculation of the ADC, the creation of ADC maps, and the determination of relative ADC (rADC) values was done as previously described.18 We showed in a previous study that the rADC values did not significantly differ between the Siemens prototype and the Vision scanner.18
Perfusion-weighted MRI.
We used a dynamic first pass bolus tracking of gadolinium diethylenetriamine penta-acetic acid (DTPA) and a multislice, gradient echo, echo-planar sequence with the following parameters: TE 60 msec, repetition time 2 seconds, FOV 260 mm, acquisition matrix 128 × 128, and the slice thickness was 7 mm with no gap between slices. The perfusion sequence usually had fewer axial slices than the diffusion sequence; thus, the imaging slices (typically 12 slices) were positioned to cover the region suspected to be ischemic based on the extent of abnormal signal intensity on the DWI. However, the selected perfusion slices were acquired at exactly the same slice position as their respective DWI counterparts. The gadolinium bolus (0.1 mg/kg body weight) and a subsequent saline flush of 20 mL were administered IV by manual injection over approximately 5 seconds. The injection was performed using an 18- or 20-gauge IV catheter in an antecubital vein. We have determined in our laboratory that manual injection by an experienced investigator is indistinguishable from that done by a power injector, which is the method of choice used currently.
Analysis of perfusion images.
The dynamic perfusion series (40 acquisitions, one every 2 seconds) was processed on a pixel-by-pixel basis to produce a variety of maps related to CBV and tissue perfusion. The injection of contrast media resulted in a T2* shortening caused by susceptibility effects with associated T2 or T2* signal loss.23-26 The changes in T2* were expressed as a change in degree of relaxation (ΔR2*, where R2* = 1/T2*) and calculated as ΔR2* (t) = [−ln(Si(t)/S0)]/TE, where ln = natural log, Si(t) = signal intensity at time (t) after injection of the contrast agent, and S0 = precontrast signal intensity. On ΔR2* images, regions of large signal change (caused by a large amount of contrast media passing through the capillary bed) have a high signal intensity, whereas regions of small signal change have a low signal intensity (figure 1). Principles of indicator dilution analysis for nondiffusible indicators were applied to the concentration-time curves. According to this theory, the area under the concentration-time curve is proportional to the local CBV.26-28 The rCBV should be better approximated as the time increases over which ΔR2* is integrated. As the time period is shortened over which the integration is done, the ΣΔR2* becomes more weighted by time and therefore represents more of a flow weighted volume image. To describe the time course of contrast through the tissue, we used a fifth order polynomial fit to calculate the tissue concentration-time curve on a pixel-by-pixel basis.
Figure 1. The relative cerebral blood volume (rCBV) calculations for three operationally defined regions (ischemic core, penumbra, and a contralateral control region [CControl]) in an individual patient. The numeric integration of the area under the curve (sum of ΔR2*) is a value that is proportional to CBV in normal brain tissue. After the tissue concentration-time course after gadolinium injection, we evaluated three different end points for our rCBV map calculations to determine which end point would maximize differences between normal and ischemic brain regions. In panel a we integrated the area under the curve up to the peak of a normal control region, and that time point was then also used as a cutoff for the operationally defined penumbra and core regions as well (initial rCBV, panel a). In panel b we integrated and calculated the area under the curve up to the peak of each of the three regions (peak rCBV), and in panel c we integrated up to the point where the tissue concentration-time curve returned to a local minimum again (total rCBV), which is referred to as the end of the first pass.
Several values were derived from these tissue concentration-time curves. Because the arterial input function was not known and was not estimated, only relative values could be derived:
-
1. rCBV: The area under the curve is the sum of ΔR2* (ΣΔR2*), which is a value known to be proportional to CBV in normal brain tissue.26 The start point for the numeric integration was that time point at which the concentration-time curve first deviated more than 0.5% from its baseline (see figure 1). Because it was our intention to find one or more indexes that would best differentiate between the operationally defined ischemic core, penumbra, and normal tissue, we chose different end points to evaluate whether we could maximize differences between normal and ischemic brain regions. We evaluated three different end points for the ΣΔR2* calculations:
(a) The area under the curve was integrated for each operationally defined region (ischemic core, penumbra, contralateral normal control region) up to the time when the contralateral control region reached its maximal susceptibility effect (figures 1a and 2⇓ ). This index is referred to as the initial rCBV.
Figure 2. This figure demonstrates horizontal slices of two different cerebral blood volume (CBV) maps (initial relative CBV [rCBV], total rCBV), the relative mean transit time (rMTT) map, and an image representing the relative cerebral blood flow (rCBF). See Methods for more details on how these images were calculated.
(b) The area under the curve was integrated for each region up to the time when each region reached its maximal susceptibility effect (see figure 1b). This index is referred to as the peak rCBV.
(c) The area under the curve was integrated up to each region’s end of the first pass (see figures 1c and 2⇑). This index is referred to as the total rCBV.
-
2. Relative mean transit time (rMTT): The MTT is the time that it takes for a bolus of paramagnetic contrast to pass through a defined volume of cerebral tissue. The rMTT (see figure 2) was calculated from the first moment to the peak in the ΔR2* fitted curves (rMTT = ∫tC(t)/∫C(t)dt). The rate of change of the concentration-time curve divided by the ΣΔR2* served as an estimate of the MTT.24,28
-
3. rCBF: Dividing the ΣΔR2* (integrated up to the peak of a contralateral normal brain region) by the rMTT provides an index of CBF (see figure 2). rCBF values are expressed in arbitrary units.
Operational definition of the penumbra.
Based on previously established and published characteristics on the ischemic penumbra,1,5,11 we developed an operational definition of tissue at risk for infarct progression using a retrospective approach. The lesion identified by the first diffusion study indicated in our operational model the ischemic core (figures 3 and 4⇓). The mismatch region between the initial diffusion lesion and its extension as demonstrated by a follow-up DWI study done 24 to 72 hours after the first one defined our operational penumbra. Using the perfusion study from day 1, indexes of CBV (initial rCBV, peak rCBV, total rCBV) and flow (rCBF), the rMTT, as well as the ADC values were determined in three defined regions (see figures 3 and 4⇓): 1) the DWI lesion at time point 1 (representing our operational ischemic core), 2) the mismatch region between the first and follow-up DWI study (representing our operational penumbra), and 3) a region on the contralateral unaffected hemisphere that grossly mirrored the first DWI lesion in extent and anatomic location. The mirror region had to be within 10% of the size of the DWI lesion at time point 1 and had to be within the same anatomic localization on the contralateral hemisphere. To exclude possible artifacts from intrasulcal or subarachnoid CSF, we manually placed these regions on all slices where they could be identified, and calculated means across slices (weighted by the number of voxels) for each of the operationally defined regions.
Figure 3. Operational model to define the ischemic penumbra. The lesion identified by the first diffusion study indicates the region of energy failure and resembles the operational ischemic core in our model. The enlargement of this diffusion lesion (in the follow-up diffusion study) into the larger hypoperfused region is the operational penumbra in this study.
Figure 4. A 72-year-old patient with a diffusion lesion enlarging into a surrounding hypoperfused region. This figure shows the two sequential diffusion studies color coded and superimposed onto the relative cerebral blood flow (rCBF) image, two cerebral blood volume images (initial rCBV and total rCBV), and a map of the relative mean transit time (rMTT). The total rCBV map demonstrates partial normalization of the CBV in the outer margins of the hypoperfused region. There was an increase in the total rCBV in the operationally defined penumbra region across the entire group.
Volumetric assessment of lesion size.
All data processing was done using Advanced Visualization Systems software (Advanced Visualization Systems, Waltham, MA) running on a Hewlett-Packard workstation (Hewlett-Packard, Palo Alto, CA). Volumes were measured on the image of maximum contrast (i.e., DWI with the highest b value) between lesion and normal brain regions. DW images as opposed to ADC maps were used for volumetric measurements due to the better contrast between abnormal and normal brain tissue. The DWI lesion and the region of ADC abnormality roughly coincided in all initial studies. Similarly, the DW image with the highest b value of the second time point was used for the volumetric assessment. A differentiation between normal and abnormal brain tissue based on the ADC maps is difficult between 3 and 4 days after stroke onset because ADC values progressively pseudonormalize.18 The b = 0 MR images, which are equivalent to T2WIs, were used for measuring the abnormal T2 volume in the chronic stage because they provide the maximum contrast between lesion and abnormal region in this stage. All our DWI and T2WI volumes are measured by at least two experienced blinded observers who measure the lesions on two occasions, and the mean value is used (intra- and interobserver reliability in measured lesion volumes was r > 0.95, with a mean deviation of less than 5% for intraobserver reproducibility).
The volume of the perfusion abnormality was only assessed on the rMTT maps because these maps provide the best contrast between a hypoperfused region and normal brain of all perfusion maps generated in this analysis. Measurements were done by one experienced observer who was blinded to the clinical scores and the DWI and T2WI lesion volumes. No volumetric measurements were done on the rCBV or rCBF maps. Volumes for the regions of interest (ROIs) drawn on the diffusion images, T2WIs, and on the rMTTs were computed by multiplying the measured area per slice by section thickness. Because there was no interslice gap, these volumes are good estimates of the full extent of the true ROI.
Statistical analysis.
For the statistical analysis, we first normalized values of the operationally defined ischemic core and penumbra to the contralateral control region. Comparisons between ischemic core and penumbra were made using a signed rank test; similar results were found with a paired t-test. Because the main focus of this study was to develop methods to characterize and separate the operationally defined penumbra from the ischemic core, we used several different approaches to developing a logistic regression model to predict the operationally defined penumbra. In the initial approach, we first used stepwise matched logistic regression to determine which of the three measures of CBV (initial rCBV, peak rCBV, and total rCBV) was the best predictor of the operationally defined penumbra. Separately, we selected the better of the two measures of flow (rCBF, rMTT). Results of these two models were compared using the Akaike information criterion (AIC); a lower value indicates a better-fitting model. Following this, we tested the improvement that other variables made to this model using a likelihood ratio rest. Because the total rCBV most closely corresponds to CBV as defined by prior PET studies, we also considered this variable as either the primary predictor or an additional predictor of region. Matched 1:1 logistic regression was used to take into account that each patient provided two measurements, one for the operational core and one for the operational penumbra region. All statistical computations were done using SAS, version 6.12 (SAS Institute, Cary, NC). A p value of less than 0.05 was regarded as significant; a corrected p value for multiple tests on the same sample is 0.0013 (0.05/39 = 0.0013).
Results.
Demonstration of the operationally defined ischemic penumbra.
The mean (± standard error of the mean [SEM]) enlargement of the diffusion abnormality across the entire group was 376.6% (±206.5) comparing the first DWI with the subsequent DWI, which was on average 48 (±6) hours later. The enlargement of the DWI abnormality occurred within the region of hypoperfusion in all the patients. All patients had a region of abnormal perfusion that was larger than the DWI abnormality at the initial scanning time point. The mean volume of abnormal perfusion (using the MTT abnormality as a marker of abnormal perfusion) was 96.6 mL (±16.9) and that of the DWI abnormality was 28.6 mL (±8.3) at the initial time point and 59.2 mL (±15.4) at the second time point. Twenty-two of 25 patients had chronic T2-weighted scans done with a mean volume of 66.6 mL (±15.9). The mean increase of the abnormal chronic T2 volume compared with the initial DWI volume across the whole group of patients was 467.0% (±182.4). Figure 4 demonstrates a representative patient of our study population with a small cortically based DWI lesion and a larger perfusion abnormality that included a major portion of the middle cerebral artery territory. The diffusion enlargement of 66% (comparing day 1 with day 2) occurred within the oligemic territory. Subtracting the DWI lesion at day 1 from the more expanded lesion of the follow-up study represents the ischemic penumbra in our operational model.
Quantitative characterization of operationally defined ischemic core and penumbra region.
Figure 5 shows the mean tissue concentration-time curve of the operationally defined ischemic core, penumbra, and a contralateral control region. The three mean regional tissue concentration-time curves showed the most pronounced differences in their initial slope. Table 1 shows the results separately for the operationally defined core and penumbra region. Both regions differed for all five measurements (p < 0.01 for all variables, Wilcoxon signed rank test). There were also differences between the operational core and a contralateral control region as well as between the operational penumbra and a contralateral control region for the initial rCBV, rMTT, and rCBF (p < 0.001). Table 2 shows the results for subgroups of patients scanned within 6 hours of stroke onset (n = 15) versus those scanned between 6 and 24 hours after stroke (n = 10). Comparing the results separately for each region, time of measurement was not an important outcome for any of the operational core measurements (all p > 0.25, Wilcoxon rank sum test), but there was some evidence for differences in the operational penumbra measurements for initial rCBV (p = 0.09, all others p > 0.2; Wilcoxon rank sum test). There were no differences over time for the differences between operational core and penumbra, however (all p > 0.20, Wilcoxon rank sum test).
Figure 5. Mean (±SEM) tissue concentration-time curves of the operationally defined ischemic core, penumbra, and a contralateral control region (CControl) derived from all 25 patients. Most pronounced differences among the three regional tissue concentration-time curves are within the first 10 seconds.
Mean relative (compared with contralateral control region) values of ΣΔR2*, rMTT, and rCBF in the operationally defined ischemic core and penumbra compared with a contralateral control (CC) region (n = 25)
Mean ± standard error of the mean relative values of ΣΔR2*, rMTT, and rCBF in the operationally defined ischemic core and penumbra compared with a contralateral control (CC) region. Patients are subdivided into two groups depending on the delay time to scanning
Using matched logistic regression, initial rCBV (p = 0.0001), peak rCBV (p = 0.0013), and total rCBV (p = 0.0001) were all significant predictors separately. The addition of either total rCBV or peak rCBV to a model including initial rCBV did not improve the model, however (both p > 0.25). Similar to the CBV measurements, rCBF (p = 0.0001) and rMTT (p = 0.0037) were also statistically significant predictors of the operationally defined penumbra, although rCBF was by far superior to rMTT in differentiating operational penumbra from core. Although both initial rCBV and rCBF were significant predictors of the operational penumbra, rCBF (AIC = 9.12) was a marginally better predictor than initial rCBV alone (AIC = 14.71; see Methods for AIC). Adding initial rCBV to rCBF led to perfect separation of the operational penumbra from core regions. Total rCBV (AIC = 20.59) was not as good a predictor as initial rCBV (AIC = 14.71). We also considered a model using both total rCBV with rCBF. The addition of total rCBV did not improve the model (p > 0.25).
The ADC in the operationally defined ischemic core and penumbra.
The ADC in the operationally defined ischemic core was reduced to 56.4% (±3.8), whereas the ADC in the operationally defined penumbra was only slightly reduced to 91.3% (±3.0) compared with a control region on the contralateral unaffected hemisphere. Anatomic regions for the ADC determination were identical to those used to determine values of blood flow and volume (see above).
Discussion.
We have shown that a combination of DWI and PWI studies can be used to operationally define cerebral tissue at risk for infarction in human stroke. The validity of the derived MR indexes of CBF and CBV to predict the ischemic penumbra has to be tested in prospective studies. The combination of both techniques has been used in differentiating the ischemic core from a penumbral region in animal studies.4,6,29-31 PWI can be used to assess regional blood supply and to delineate a region of decreased perfusion that may or may not proceed to infarction. DWI is a sensitive marker of energy-requiring ionic membrane gradients, and the viability of the penumbral tissue requires maintenance of energy-dependent processes,5 making DWI an ideal tool for defining ischemic tissue in the hyperacute stage of human stroke. The acute reduction in ADC is associated with a depletion of high-energy compounds30-32 and failure of energy-requiring processes,5,20-22 indicative of the ischemic core region.
The applied operational model that we used to define and quantitatively characterize the ischemic penumbra made the assumption that the DWI abnormality at time point 1 describes the zone of energy failure and as such an operationally defined ischemic core within a larger region of oligemia as identified by the initial perfusion study. This assumption is supported by animal studies.4,6,29-31 Recovery of the DWI abnormality has only been shown in animal experiments after a short temporary occlusion, with ADC values only moderately reduced.33 Further support for this assumption comes from the chronic T2WIs that were obtained in 22 of 25 patients. In all these patients, the T2WI lesion was larger than the DWI lesion at time point 1, which argues against the notion that the DWI enlargement could just be due to edema. The enlargement of the initial diffusion lesion (as demonstrated by the second DWI study) into the oligemic region was used to define cerebral tissue at risk that can evolve into infarction over time and as such defined the operational penumbra in this study. In using anatomic templates of our operationally defined ischemic core and penumbra, and overlying them on various synthesized maps of CBF and CBV as well as onto maps of the ADC, we found that the initial rCBV and the rCBF were significant predictors of the operationally defined penumbra. If the MRI-derived values are confirmed in future prospective studies, they can be used to determine the presence of tissue at risk for infarction in the hyperacute stage of ischemic stroke.
We used normalized data in this study as a way of reducing the intersubject variability. This approach may be to an underestimation of the actual abnormality in cases with significant ipsilateral arterial disease that could potentially influence some of our perfusion parameters. Only one of our patients showed significant ipsilateral arterial disease, whereas several others showed mild ipsilateral atherosclerotic disease demonstrated on magnetic resonance angiography. MR indexes of the patient with significant ipsilateral arterial disease did not differ from all other patients’ values.
Among the various estimates of the rCBV, the initial rCBV was the best predictor of the operational penumbra. The initial rCBV reflects the area under the curve for the initial part of the concentration-time curve and is therefore an index of the rCBV that is weighted toward the volume of inflowing blood. The initial rCBV is very similar to the rCBF in this study, which can be seen in figure 2 and in table 1 showing similar proportional differences between the operationally defined penumbra and ischemic core. Although the initial rCBV was significantly reduced, the total rCBV was significantly increased in the operational penumbra compared with a contralateral control region. However, the total rCBV was not as good as the initial rCBV in differentiating the operational penumbra from the ischemic core. The increase of the total rCBV in our operational penumbra concurs with most PET measurements showing an increase in CBV in the ischemic penumbra caused by an early compensatory response to reduced cerebral perfusion pressure.12,13 The rCBF was slightly better than the initial rCBV in differentiating the operationally defined penumbra from core. The rMTT was significantly prolonged in the operational penumbra and even more in the ischemic core. However, the rMTT was not as good as any of the other markers in differentiating the operationally defined penumbra from ischemic core but was an excellent marker in differentiating abnormal (i.e., core and penumbra) from normal perfusion (see table 1).
Our operationally defined penumbra did show a marginal but significant reduction in the rADC on day 1 compared with an unaffected brain region. The slight drop in the mean regional rADC was not enough to cause a signal increase in the diffusion image, which is a well-described phenomenon. According to Kohno et al.,32 only CBF values of less than 40 mL/100 g/min yielded increases in signal intensity in DW images. This argues that CBF has to be reduced by more than 20% from a presumed mean cerebral perfusion value of 50 mL/100 g/min34 to be associated with a significant change in the ADC. Furthermore, a reduction of the rADC in our operational penumbra by 8% (as seen in our study) does not significantly interfere with the maintenance of transmembranic ionic gradients.5 In animal studies it has been shown that the region of ADC reduction of less than 80% compared with that of a normal brain region correlates well with the histologically defined region of infarction.31 The marginally reduced rADC in the operational penumbra in our study may be due to intermittent spreading depression,35 which is associated with lesion growth.10 Spreading depression has been demonstrated in animal stroke models, and transient ADC declines were interpreted as correlates of peri-infarct depolarizations.35 Alternatively, the rADC decrease could also reflect the presence of a small subset of neurons with impaired ionic membrane gradients accounting for an overall slight reduction of rADC values within our operational penumbra.
Functional data on validated thresholds of CBF, CBV, or oxygen consumption to distinguish the ischemic core from the penumbra and oligemic normal tissue in human stroke are rare.2,3,5,7,8 The classic definition of the penumbra is that of a region evolving over time toward infarction with CBF values below those needed to sustain electrical activity but above those required to maintain cellular ionic gradients.1,5 We have learned from human PET studies that untreated ischemic regions with very low CBF (<10 mL/100 g/min) rapidly convert into irreversibly damaged tissue and constitute the ischemic core.2,3,9 Surrounding, adjacent, or intermixed zones of less severely impaired CBF (within 10 to 20 mL/100 g/min, although some reports give a range between 7 and 45 mL/100 g/min) constitute the ischemic penumbra.1,5,7,8,22,36 Our data are in good agreement with CBF thresholds derived from PET studies. The mean CBF of a mixed gray/white matter normal human brain region is approximately 50 mL/100 g/min34; a reduction of our calculated CBF index, rCBF, in our operationally defined penumbra to 37% of normal values would translate into a CBF value of 18.5 mL/100 g/min. Similarly, the reduction to 12% of normal values in our operationally defined ischemic core would translate into a CBF of 6 mL/100 g/min.
There is some overlap in these threshold values as has been shown in several recent PET studies,7,8 and this is also evident from our data (see table 1). However, there was less overlap between rCBF or initial rCBV values in the operationally defined penumbra if the MRI studies were performed less than 6 hours after stroke onset (see tables 1 and 2⇑). The greater overlap in MRI values between the operational penumbra and ischemic core at later time points (MRI studies done more than 6 hours after stroke onset) could be due to partial reperfusion of the core region or the continued deterioration of the operational ischemic penumbra, or both. The duration of reduced CBF or that of other markers that might trigger irreversible cell damage and determine the point of no return are still unknown.3,5 It is known that the penumbra does not remain viable for extended periods, and there is experimental evidence that CBF thresholds increase with increasing duration of the ischemia.3,5 Animal stroke models have shown that the infarct core (in which ATP is depleted) gradually expands into the ischemic penumbra over a few hours.3,4,6
One might challenge our assumption that the entire region of diffusion abnormality represents the ischemic core. Indeed, the DW image itself may be misleading, and one has to calculate maps of the ADC for the objective interpretation of the diffusion abnormality. We have learned from our previous experience18 that the mean rADC within the hyperintense DWI lesion is on average reduced to approximately 58% of control values in the acute stage of an infarction. It is known from animal experiments that this reduction corresponds to a CBF of 10 to 15 mL/100 g/min and that this is associated with failure of the energy requiring Na-K-ATPase.5,14 In animal stroke models it has been suggested that there may be a small outer rim within the DWI lesion that has characteristics of a penumbral region revealing acidosis but no depletion of ATP initially, but over a period of 2 hours the region of the ATP depletion enlarges22,32 and approximates the entire region of diffusion abnormality. Current technologies and image resolution used in human studies make it very difficult to discern such a small outer rim within the initial DWI lesion. Nevertheless, there is a good correlation between the animal data applying this model and our study. Hoehn-Berlage et al.30,31 showed that the region of energy depletion (ischemic core) had CBF values of 18 ± 14 mL and rADC values of less than 77 ± 3% compared with the contralateral side. The region of tissue acidosis (resembling the ischemic penumbra) had flow values of 31 ± 11 mL with rADC values below 90 ± 4%.
Quast et al.6 used a model similar to ours in which a DWI lesion (ischemic core) was differentiated from a larger perfusion lesion surrounding the DWI lesion (ischemic penumbra). The majority of human strokes have a mismatch between a smaller diffusion and a larger hypoperfusion lesion. The diffusion lesion grows into the larger and surrounding hypoperfused region but rarely reaches its size.15,37-40 Thus, just simply using the mismatch between the diffusion and perfusion lesion to operationally define the penumbra may mix two subregions that are different, one region that is at risk for infarction and may infarct (our operationally defined penumbra) and another region that is at risk but spontaneously escapes (hypoperfusion beyond the operationally defined penumbra). In the present study, we wanted to focus on that part of the ischemic penumbra that evolves into infarction because acute therapeutic interventions should be most helpful for this particular tissue compartment. DWI and PWI permit the definition of irreversibly damaged and potentially salvageable hypoperfused tissue.15,39-42 This is particularly true in the first hours after stroke onset, when prediction and identification of potentially salvageable tissue is of most clinical relevance.
Acknowledgments
Supported by grants from the National Institute of Neurological Disorders and Stroke, the American Heart Association, Harcourt General Charitable Foundation, and by the Phyllis and Paul Fireman Fellowship. A.E.B. is supported by the National Health and Medical Research Council of Australia. K.O.L. was a grantee of the Swiss National Science Foundation.
Acknowledgment
The authors thank Dr. L. Caplan for helpful suggestions.
Footnotes
-
Preliminary results were presented at the 49th annual meeting of the American Academy of Neurology; Boston, MA; April 1997.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.