Refining the Mismatch Concept in Acute Stroke: Lessons Learned from PET and MRI

Abstract

In ischemic stroke, positron-emission tomography (PET) established the imaging-based concept of penumbra. It defines hypoperfused, but functionally impaired, tissue with preserved viability that can be rescued by timely reperfusion. Diffusion-weighted and perfusion-weighted (PW) magnetic resonance imaging (MRI) translated the concept of penumbra to the concept of mismatch. However, the use of mismatch-based patient stratification for reperfusion therapy remains a matter of debate. The equivalence of mismatch and penumbra, as well as the validity of the classical mismatch concept is questioned for several reasons. First, methodological differences between PET and MRI lead to different definitions of the tissue at risk. Second, the mismatch concept is still poorly standardized among imaging facilities causing relevant variability in stroke research. Third, relevant conceptual issues (e.g., the choice of the adequate perfusion measure, the best quantitative approach to perfusion maps, and the required size of the mismatch) need further refinement. Fourth, the use of single thresholds does not account for the physiological heterogeneity of the penumbra and probabilistic approaches may be more promising. The implementation of this current knowledge into an optimized state-of-the-art mismatch model and its validation in clinical stroke studies remains a major challenge for future stroke research.

Introduction

Neuroimaging has become a key technology in stroke research. The current portfolio of clinical imaging modalities, e.g., multimodal computed tomography, stroke magnetic resonance imaging (MRI), single-photon emission computed tomography, and positron-emission tomography (PET) contributes to an in-vivo insight into stroke pathophysiology. First, PET provided detailed metabolic characteristics of ischemia (Astrup et al, 1981; Baron et al, 1981). It then took decades until MRI was available for acute stroke diagnosis (Warach et al, 1995). This improvement in imaging stimulated therapeutic innovations. Stroke was pulled out of the diagnostic and therapeutic shadows and the concept of early therapy was implemented in practice and included in evidence-based guidelines (Lees et al, 2010).

The large body of experimental stroke research shows that the ischemic damage has two major predictors: the severity and the duration of ischemia (Heiss and Rosner, 1983; Jones et al, 1981). Ischemia leads to an impairment of metabolic and electrophysiological properties. According to the characteristic of the ischemic stress, this impairment may be reversible in case of early reperfusion, or may become irreversible according to the severity of ischemia and the time without adequate reperfusion. The tissue with an uncertain destiny—showing preserved viability but penumbral perfusion and mostly surrounding the infarct core—was termed as ‘penumbra.’

The experimental description of penumbra was first translated to human stroke by PET (Baron et al, 1981; Heiss et al, 1999; Lenzi et al, 1982). Although of high pathophysiological specificity, the logistic needs of PET imaging do not allow routine clinical application. The introduction of diffusion-weighted (DW) and perfusion-weighted (PW) MRI thus opened a new era of stroke imaging and a second translation was realized: from the PET-based concept of penumbra to the MRI-based concept of mismatch (Warach et al, 1995). Despite the undisputed improvements in stroke imaging with MRI, and despite very promising data from clinical studies, the evidence for the mismatch-based patient stratification, e.g., for thrombolysis, is still lacking (Mishra et al, 2010). The reasons for this ‘translational barrier’ are manifold and are the target of stroke research to improve the mismatch concept.

The following overview will highlight the current evidence of mismatch imaging in human stroke, focusing on PET and MRI. It will discuss important methodological issues and will identify needs for future imaging research.

The positron-emission tomography based concept of penumbra

The pivotal event of ischemia is the decrease of cerebral blood flow (CBF). Positron-emission tomography with 15O-water and with 15O-CO₂ using arterial blood sampling allows quantitative measurement of regional flow values. This method is considered as the gold standard of CBF imaging. For human stroke, relevant flow thresholds have been identified as they may predict tissue fate in the acute phase. However, the definition of a single flow threshold in ischemia remains difficult because of several influential factors: (1) there is interindividual and age-dependent variability of CBF; (2) the thresholds obtained differ between white and gray matter (Pantano et al, 1984); (3) region of interest-based analyses yield different results compared with voxel-based analyses (Marchal et al, 1999); and (4) the duration of ischemia is an independent predictor of tissue outcome (Jones et al, 1981). In this respect, the CBF threshold that defines critical ischemia (leading to irreversible infarction) ranges from 8 to 12 mL/100 g per minute (Furlan et al, 1996; Heiss et al, 1998; Marchal et al, 1999). Tissue experiencing perfusion values below this threshold has a high probability to turn into irreversible infarction unless perfusion is quickly restored. The CBF values >20 mL/100 g per minute, on the other hand, do not lead to a functional impairment of neuronal tissue and are termed as oligemia or normoperfusion. Most important, CBF values between the irreversibility threshold and 20 mL/100 g per minute describe the ‘penumbral flow.’ Tissue with perfusion values within this range experiences a functional impairment but remains viable for a certain time (Baron et al, 1981; Heiss et al, 1999; Lenzi et al, 1982).

Perfusion imaging, however, represents a ‘snapshot’ of the dynamic process of ischemia. In contrast to animal models, perfusion imaging in human stroke cannot control for onset of ischemia or perfusion changes before imaging. For example, an initial critical ischemia (i.e., CBF values <8 mL/100 g per minute) might show a partial reperfusion (i.e., CBF values around 20 mL/100 g per minute) at the time of the imaging. Thus, the ‘snapshot’ might be misleading as the tissue may have already lost its viability. An additional parameter of tissue viability is therefore needed to better explain the tissue outcome. For PET, this ‘missing link’ is the oxygen metabolism imaged by 15O-O₂ PET (cerebral metabolic rate of oxygen) (Baron et al, 1981, 1984; Baron, 1999; Heiss, 2000; Marchal et al, 1999). Values <65 μmol/100 g per minute indicate irreversible tissue loss regardless of perfusion values. Values above this threshold indicate viable tissue even with concomitant hypoperfusion. This metabolic threshold is method dependent. It refers to large cortical regions of interest and is lower in a voxel-based analysis (Marchal et al, 1999). Put simply, the balance of CBF and cerebral metabolic rate of oxygen, i.e., the oxygen consumption per unit of oxygen supply, is expressed by the oxygen extraction fraction (OEF). This unique PET measure represents the percentage of extracted oxygen from the blood supply. Whereas the brain under physiological circumstances shows OEF values of ∼30%, ischemic challenges are compensated by OEF increases up to 80%. This elevation of OEF, together with flow between the penumbra and viability thresholds, is the in-vivo hallmark of the penumbra. Since the OEF measurement requires a bolus or steady-state inhalation approach with complex logistics, alternative techniques have been successfully described to identify the penumbra. Among them, 11C-flumazenil (FMZ) and 18F-fluoromisonidazol (FMISO) are the most promising. 11C-FMZ is a marker of cortical neuronal integrity and may be useful for delineation of the penumbra in combination with 15O-water PET (Heiss et al, 2004). 18F-MISO is a marker of hypoxic tissue that may allow direct visualization of hypoxic impairment (Markus et al, 2002) but may have some limitations (Takasawa et al, 2008b).

It has to be kept in mind that cerebral ischemia is a dynamic process with an individual development of penumbral tissue (Heiss and Rosner, 1983). Thus, imaging findings have to be interpreted with respect to the time frame, the individual ischemic patterns and the degree of collaterals. This becomes highly relevant if imaging-based time windows for acute therapy are targeted.

The magnetic resonance imaging-based concept of mismatch

The definition of mismatch applies a two compartment approach: the infarct core is delineated on maps of DW intensity or of the apparent diffusion coefficient, the area of hypoperfusion is delineated on maps of PW imaging (Baird et al, 1997; Parsons et al, 2002; Schellinger et al, 2001). The volumetric difference between these two compartments, i.e., the tissue with normal appearance on DWI but hypoperfused on perfusion-weighted imaging (PWI), is termed as ‘mismatch.’ The mismatch is considered as ‘at risk’ of infarct growth without reperfusion, and shows characteristics of the penumbra. This concept has been substantiated by several imaging studies using PET and MRI. It was shown that the core volume correlates with stroke severity and predicts large parts of the finally infarcted tissue. It was also shown that, in addition to the core, the mismatch/penumbra contributes to the neurologic deficit (‘symptomatic tissue,’ i.e., core plus penumbra). Finally, it was shown that the rescue of penumbral tissue correlates with clinical recovery (Alawneh et al, 2011; Furlan et al, 1996; Heiss et al, 1998; Marchal et al, 1999; Muir et al, 2006). Although this concept seems straightforward, the equation mismatch=penumbra has to be considered carefully as several methodological issues remain unresolved (Alawneh et al, 2011; Donnan and Davis, 2002; Kidwell et al, 2003; Sobesky et al, 2005) and as clinical studies of acute stroke therapy that used the mismatch concept for patient stratification produced inconclusive results (Mishra et al, 2010). Thus, the current mismatch concept has to be evaluated with respect to the following questions: (1) Can DW identify the infarct core? (2) Can PWI identify penumbral flow? (3) Can the mismatch reliably identify the penumbra? These questions can be addressed by two types of studies: on the one hand, by magnetic resonance (MR) studies with serial follow-up; here, the analysis of tissue patterns shows which pattern may define irreversibly infarcted or salvageable tissue in the presence or absence of reperfusion. This approach yields large patient numbers but experiences unknown flow changes between early and late imaging. On the other hand, these questions can be addressed by a validation of MR findings with a reference method as for example PET. This validation is based on small patient samples but allows a direct comparison of hemodynamic and metabolic findings with respective PET results. However, if this comparative imaging is not performed simultaneously, this approach experiences the uncertainty of possible physiological changes between the scans which need to be taken into account.

Imaging of the Infarct Core

Since the first description of DW signal alterations after human stroke (Warach et al, 1995), DWI remains the fastest clinical assessment of ischemic changes. The DW lesion is a surrogate of tissue injury as it detects the Brownian motion of water molecules within the interstitial space. Ischemia leads to cell swelling and to a consecutive decrease of the intercellular space. The resulting restriction of Brownian motion is depicted as a signal alteration on DWI.

A close correlation of early DW lesion and final infarct volume was seen in early pivotal MR studies (Warach et al, 1995). This finding initiated an extensive evaluation of DW imaging to show which percentage of the finally infarcted tissue can be predicted in the acute phase of stroke and which percentage of the acute DW lesion represents inevitable infarction (Kranz and Eastwood, 2009). Combined PET/MRI studies, comparing the DW signal alteration with PET markers of tissue viability, have shown that both, MRI and PET, may predict up to 85% of the final infarct. However, DWI lesions with normal appearance on FMZ-PET imaging were found in 25% of the cases and escaped infarction (Sobesky et al, 2005). This pattern of false positive DW lesion is well explained by the finding that the DWI lesion includes areas of preserved tissue viability and may show penumbral patterns (Guadagno et al, 2004). These PET findings were supported by numerous MR studies that found DW lesion to be false positive in an average of 24% of the cases (Kranz and Eastwood, 2009) and single studies reported DWI/apparent diffusion coefficient reversal in ∼30% to 50% of the patients/voxels after reperfusion (Carrera et al, 2011a; Fiehler et al, 2002; Kidwell et al, 2002; Olivot et al, 2009a). The inclusion of flow surrogates, as for example the mean transit time (MTT) or Tmax, further improved the prediction of final infarction not only outside, but also within the acute DW lesion (Carrera et al, 2011a; Olivot et al, 2009a). Although DW lesion reversal has been proven in several imaging studies its relevance for clinical decision making remains unclear. The DWI reversal is difficult to predict and its resulting effect on the mismatch classification as well as on the true tissue salvage remains a matter of debate (Campbell et al, 2012; Chemmanam et al, 2010). The current data thus indicate that DW reversal is associated with hyperacute imaging, with DWI lesions of low intensity, with only moderate perfusion deficits and with early reperfusion. Its clinical relevance in terms of patient stratification remains to be clarified. Diffusion-weighted imaging should therefore be interpreted as an infarct marker with high sensitivity (which is clinically desirable) but with lower specificity (which is relevant for outcome studies).

The delineation of ‘DWI-positive’ lesions is mostly performed by visual analysis. It has to be kept in mind that DW alterations show continuous values and that their appearance strongly depends on the displayed image window. A threshold-based definition of DW lesion is clearly preferable and several studies suggest a relative DW intensity of 120% as the optimal cutoff value for final infarct prediction (Heiss et al, 2004; Na et al, 2004). However, the dichotomization into ‘DWI normal’ versus ‘DWI abnormal’ using defined cutoff values—although helpful for comparison of imaging studies—is a simplified view of the penumbra concept. Taking into account the dynamic nature of ischemia, the DW signal alteration is a function of both the severity and the duration of hypoperfusion (Jones et al, 1981). It is therefore unlikely that a single DW threshold will adequately differentiate core from penumbra for all time points and individual hemodynamic patterns. Accordingly, comparative MR/PET studies found that parts of the DWI lesion displayed penumbral patterns and were reversible (Fiehler et al, 2002; Guadagno et al, 2004; Sobesky et al, 2005). This emphasizes that a comprehensive approach to DW values should be targeted in future studies.

Imaging of Hypoperfusion

Perfusion-weighted imaging has been shown to be the ‘Achilles’ heel’ within mismatch concept: according to the classical mismatch definition, the PW-based flow values alone determine the presence or absence of mismatch within the tissue defined as ‘DWI normal.’ This leads to an estimate of penumbral tissue by MRI if the equation ‘mismatch=penumbra’ is assumed. Perfusion-weighted imaging uses a bolus tracking technique after application of a paramagnetic intravascular contrast agent. To analyse the flow patterns, different curve parameters can be derived with or without a deconvolution using an arterial input function (AIF) from large brain vessels (Ostergaard et al, 1996). The AIF allows only an approximation of the true plasma concentration of the contrast agent, due to characteristics inherent to T2^∗ imaging (Ostergaard et al, 1996).

It is evident that PWI differs substantially from the PET-based CBF measurement where the activity of a partly diffusible tracer (15O-H₂O) is measured using an AIF derived from the radial artery (Herscovitch et al, 1987). The use of PWI for blood flow measurement includes several important steps:

(1) Choice of the adequate parameter map: Maps of time-to-peak (TTP) as a simple measure of arrival time of the contrast agent do not represent the complex hemodynamic principle of CBF, but provide a robust estimate of hypoperfusion in acute stroke trials. Maps of TTP, if normalized to the unaffected hemisphere (relative TTP, rTTP) are well comparable between individuals as well as between different imaging facilities and do not rely on time-consuming postprocessing (Neumann-Haefelin et al, 1999; Sobesky et al, 2004; Zaro-Weber et al, 2010c). However, the delay of tracer arrival is not necessarily accompanied by decreased absolute CBF, as for example in chronic carotid occlusion, and may render TTP values false positive and susceptible to collateral circulation (Yamada et al, 2002). For a better representation of hemodynamic parameters, maps of TTP, cerebral blood volume, MTT, and CBF are calculated by a deconvolution procedure using an AIF (Ostergaard et al, 1996). Currently of major clinical interest, maps of Tmax, which represent TTP after deconvolution, are widely used and implemented in recent multicenter stroke studies (Albers et al, 2006; Davis et al, 2008). Maps of MTT are surrogates of hypoperfusion since they show the increase of transit time in early ischemia and have been described in many clinical studies (Butcher et al, 2005; Olivot et al, 2009b; Schellinger et al, 2001; Thijs et al, 2001). In comparative MR/PET studies, maps of MTT gave a fair estimate of penumbral patterns and contributed to the performance of DWI infarct prediction (Carrera et al, 2011a,2011b). Maps of MTT and TTP allow a good visual delineation of signal alteration. Due to the low contrast between gray and white matter these maps appear homogeneous and are preferred for visual and volumetric analysis. Maps of cerebral blood volume represent the degree of vascular dilatation/constriction and show increasing values in mild-to-moderate ischemia but decreasing values in the beginning of infarction. They are thus used in combination with maps of CBF to specify tissue impairment (Grandin, 2003). Maps of CBF have been tested in small single center studies (Rivers et al, 2006) but have not yet been implemented in multicenter studies.

To date, there is no consensus which PW map best identifies hypoperfusion and predicts infarct growth or response to thrombolysis (Kane et al, 2007b). Deconvolved maps are considered as superior for the detection of hypoperfusion from a theoretical point of view. However, their superiority has not yet been proven in clinical studies (Christensen et al, 2009; Grandin et al, 2002; Zaro-Weber et al, 2009).

(2) Application of thresholds: Maps of TTP, MTT, and CBF may yield comparable information about the presence and site of perfusion impairment if analyzed in a qualitative manner. To differentiate penumbral flow from benign oligemia, however, a visual analysis is not adequate and thresholds are necessary. Clinical stroke studies, aiming at the area of infarct growth, described the following range of thresholds: rTTP: >2 to >8 seconds; rMTT: >4 to >8 seconds; CBF: <20 to <42 mL/100 g per minute; Tmax: >2 to >8 seconds (Grandin et al, 2002; Liu et al, 2000; Neumann-Haefelin et al, 1999; Olivot et al, 2009b; Rivers et al, 2006). Comparative MR/PET studies showed a large variability of the calculated mismatch volume according to the applied thresholds for maps of TTP and CBF, mostly overestimating the penumbra (Sobesky et al, 2004; Zaro-Weber et al, 2009). As the congruence of mismatch and penumbral tissue was incomplete in these studies, the optimal threshold for PWI maps remained a compromise between underestimation and overestimation of penumbral flow and had to be adapted according to the desired sensitivity and specificity. A comparative PET/MR study in 26 acute stroke patients validated PW maps with respect to their performance (area under the curve) to identify penumbral flow (<20 mL/100 g per minute) and yielded the following best thresholds: rTTP: 4.2 seconds (area under the curve: 0.94); MTT 5.3 seconds (0.86); CBF 21.7 mL/100 g per minute (0.92), and Tmax: 5.5 seconds (0.94) (Zaro-Weber et al, 2010b,2010c). These results were in agreement with a PET/MR study of five stroke patients that described a penumbra threshold of 4.8 and 5.4 seconds for rTTP and Tmax (Takasawa et al, 2008a). On the one hand, these are promising results that show a good performance of maps of rTTP, Tmax, and CBF. On the other hand, there was a substantial interindividual variation that attenuated the reliability of mismatch detection in individual patients.

Apart from the improvement of PW-based flow measurement, it has to be kept in mind that penumbral flow thresholds are mainly probabilistic due to differences in metabolic needs among gray- and white-matter areas, effects of age and several other factors. For reasons of feasibility, this probabilistic approach has not yet been implemented in therapeutic trials.

(3) Quantification of PW images: The quantification of PW maps remains a major source of heterogeneity. The type of deconvolution (Ostergaard et al, 1996; Wu et al, 2003) as well as the placement of the AIF (Ebinger et al, 2010a; Thijs et al, 2004) lack a standardized definition across the imaging facilities. The MRI studies (Kane et al, 2007a), comparative MR/PET studies (Sobesky et al, 2004; Takasawa et al, 2008a; Zaro-Weber et al, 2010b) and Xenon computed tomography studies (Olivot et al, 2009b) showed that these different approaches to PW maps cause a high variability of flow values with a substantial influence on the resulting mismatch volumes. To attenuate this variability, calibration techniques for PW maps were used and could improve the reliability of perfusion measurement (Ostergaard et al, 1998; Takasawa et al, 2008a; Zaro-Weber et al, 2010a). It is important to emphasize that even deconvolved maps benefit from a calibration for global flow using the mean value of the unaffected hemisphere. This may explain why nondeconvolved maps, as for example maps of rTTP, are not inferior compared with deconvolved maps in acute stroke imaging trials. It is assumed that the use of a contralateral reference region confers a ‘composite’ of the complex local tracer characteristics and corrects for the variability inherent in the method (Christensen et al, 2009; Zaro-Weber et al, 2010b). However, the efforts to increase the precision of PWI to estimate CBF should not distract from the fact that perfusion is a physiological parameter with a high inherent variability. The predictive power of PWI (and partly of DWI as well) changes with the time from stroke onset, the site of tissue imaged (e.g., gray matter versus white matter), the degree of reperfusion, and the patient age (Bristow et al, 2005; Falcao et al, 2004; Wu et al, 2006). Thus, a voxel-based probabilistic approach incorporating imaging parameters (e.g., values of DWI, TTP, CBF, cerebral blood volume, and MTT) as well as nonimaging parameters (e.g., age, time from stroke onset, and site of tissue) might better reflect the complexicity of a predictive model (Wu et al, 2006) and has to be evaluated for its clinical feasibility.

(4) Choice of the postprocessing software: many of the procedural differences discussed above have a basis in the different postprocessing software packages that are used. They require different degrees of observer interaction and use different postprocessing methods, which renders a comparison between the studies difficult. A recent comparison of various perfusion software packages found substantial differences in the calculated volume of hypoperfusion for maps of MTT, CBF, and Tmax among three tested software packages (Galinovic et al, 2011a). Acute stroke treatment requires a fast assessment of mismatch and the use of an automated software with low processing time and few observer interaction is desirable. In this respect, an automated patient selection software (RAPID) that was used for the reanalysis of the MR data sets from two large thrombolytic trials brought encouraging results. The major outcome patterns were correctly identified on MR images by this software (Lansberg et al, 2011). This issue is of high clinical relevance and needs further evaluation in stroke trials.

Definition of Mismatch

The volume of mismatch as the amount of salvageable tissue mainly depends on the definition applied. Without standardization of mismatch calculation, the current evidence remains inconclusive and is based on differing study designs (Kane et al, 2007b; Rivers et al, 2006). Comparative PET/MR studies that validated a common mismatch definition (rTTP >4 seconds for PWI; DWI lesion threshold of >120%) found that the mismatch overestimated the penumbra and included oligemic tissue. In clinical terms, this would downgrade a possible MR-based stratification and would include patients with oligemic tissue that is not at risk. Increasing the TTP threshold (i.e., including only more severely affected tissue) partially improved the results (Sobesky et al, 2004, 2005). Furthermore, an adequate calibration clearly improved the detection of penumbral flow (Zaro-Weber et al, 2010a,2010c). As for the DWI lesion, there is no consensus about the minimal volume required. Reports of patients presenting acute PWI lesions but no DW lesion underline that PWI maps alone may define the mismatch (Cho et al, 2009). An issue of major importance is the percentage of mismatch required for therapeutic decisions. It is unclear which mismatch volume justifies aggressive therapy, since the best mismatch ratio for stroke trials remains unclear. The arbitrary volumetric difference of 1.2 (DWI lesion:PWI lesion) that was used in many studies for patient stratification has not yet been evaluated. In retrospective analyses, however, mismatch ratios up to 2.6 have been described to predict the optimal response to thrombolysis (Kakuda et al, 2008). Finally, a precise coregistration of DW and PW images has been shown to be of considerable importance to evaluate different mismatch ratios (Ma et al, 2009; Nagakane et al, 2011).

The evaluation of the mismatch concept thus may differ according to the following assumptions:

(1) The concept of mismatch is valid but needs a refinement of blood flow measurement by PWI: This refers to respective studies that improve postprocessing of PW maps. An optimized and standardized use of PWI—probably available in the near future—should in all likelihood enable a high congruence between mismatch and penumbra for clinical studies.

(2) The classical mismatch cannot reliably identify the penumbra since a single flow threshold cannot represent the complexity of penumbral pattern: This statement refers to the fact that the concept of penumbra describes a continuous spectrum of cerebral metabolic rate of oxygen and CBF values, whereas the concept of mismatch dichotomizes DW images into ‘normal’ and ‘abnormal’ and adds one PW-based cutoff value. This simplified approach does not account for the regional balance of flow and oxygen metabolism and causes a blurred version of the penumbra pattern.

Mismatch: the clinical evidence or ‘lost in translation’?

There is an important heterogeneity in the use of the mismatch concept with high impact on the assessment of the tissue at risk. This heterogeneity includes the mismatch definition itself (Kane et al, 2007b), the choice of the PW map and the method of postprocessing (Kane et al, 2007a), the use of an automated processing software (Galinovic et al, 2011a), the delineation of core and hypoperfusion (Ay et al, 2008), and the predefined mismatch ratio (Kakuda et al, 2008). If the different definitions of mismatch would be applied to one patient sample, the variability in resulting mismatch volumes would clearly exceed 20% (Kane et al, 2007a)—more than the operationally defined minimum mismatch ratio required for therapeutic decisions (mismatch ratio 1.2) according to previous stroke studies (Albers et al, 2006; Davis et al, 2008).

Despite these unsolved issues, retrospective observational studies have shown that delayed thrombolysis can be performed safely if patients are selected by a (yet poorly standardized) mismatch (Schellinger et al, 2007). This finding was promising but does not provide adequate evidence. There are a large number of single center studies in comparison with the few randomized controlled trials (Albers et al, 2006; Davis et al, 2008; Hacke et al, 2009). A recent meta-analysis of these trials concludes that a mismatch-based delayed thrombolysis cannot be recommended as part of routine care (Mishra et al, 2010). However, from a present day perspective, the studies differed in relevant conceptual issues: in the choice of thrombolytic treatment (recombinant tissue plasminogen activator versus desmoteplase); the imaging stratification for treatment (noncontrast computed tomography versus mismatch imaging); the delineation of the DW lesion (visual versus threshold); the definition of hypoperfusion (visual versus threshold Tmax >2 seconds); the choice of the PW map (maps of Tmax versus free choice); the magnitude of mismatch ratio (geometric 1.2 ratio versus visual impression of 1.2 ratio); and the presence of a placebo group (yes versus no). Considering the current knowledge of mismatch imaging, none of the studies has used an adequate flow threshold. Therefore, the conclusion of the meta-analysis is consistent with respect to the available clinical data but should not be misinterpreted as a general rejection of the mismatch concept. Instead, it should be interpreted as an evaluation of the current use of the mismatch concept and as a challenge to improve, validate, and standardize stroke MRI in the future.

Modification and extensions of the mismatch concept

To simplify or to enhance mismatch imaging, several modifications are of current interest:

The clinical/DWI mismatch combines the clinical stroke severity with the size of the DWI lesion to estimate the amount of mismatch (Davalos et al, 2004). The clinical deficit results from tissue impairment within either the mismatch or the infarct core. Thus, the stroke severity compared with the DWI lesion approximates the mismatch volume. Although this concept is a valuable tool, especially when PWI is not available, it does not reach the specificity of the DW/PW concept (Davalos et al, 2004).

The magnetic resonance angiography/DW mismatch combines the DWI lesion size with magnetic resonance angiography and thus reduces the flow information of PWI to the presence or absence of a vessel occlusion (Lansberg et al, 2008). This approach is tempting since it avoids the difficulties and controversies of PW imaging. Additionally, it focuses on a high-risk population of stroke patients which might be of interest in clinical studies (Galinovic et al, 2011b). However, it does not identify distal occlusions and does not account for collateral flow, regional perfusion, or penumbral patterns which are major determinants of tissue outcome.

Of considerable current interest, the fast fluid-attenuated inversion recovery (FLAIR)/DWI mismatch has been proposed for patients with unclear stroke onset (mostly as ‘wake up stroke’) (Thomalla et al, 2009). The DWI lesion is seen immediately after ischemia, the FLAIR lesions appear several hours after stroke. Thus, the FLAIR/DWI mismatch suggests that if a DW lesion is visible but FLAIR imaging is negative, then the probability of an onset of ischemia <4.5 hours is very high. In conjunction with mismatch imaging FLAIR imaging may help to identify patients that can be treated by thrombolysis when the onset of ischemia is unknown (Aoki et al, 2010; Ebinger et al, 2010b).

This link between imaging finding and time window points toward a major issue of mismatch research: if the MR patterns of tissue at risk are well defined, then a time-based treatment window might be replaced by an imaging-based treatment window (‘tissue clock’). The current evidence for intravenous thrombolysis within 4.5 hours is mainly based on noncontrast computed tomography, stroke severity and time after stroke onset, and thus summarizes different stroke etiologies and patterns of penumbra (Lees et al, 2010). This time window does not represent a sharp biological cutoff but is a statistical compromise between treatment risk and benefit in patients drawn from a heterogeneous population. Considering the individual variability in duration of penumbra or mismatch (Giffard et al, 2004; Gonzalez et al, 2010) an individual time window based on imaging findings seems preferable (Baron et al, 1995).

It has to be kept in mind, however, that the identification of tissue at risk is only one of many predictors of therapeutic success. The treatment-related risks have to be considered and limit the therapeutic options, e.g., the risk of intracerebral hemorrhage with thrombolysis. The mismatch concept therefore has to be extended to the assessment of total DWI lesion size, of white matter lesion load, of extensive microbleeds, and of the integrity of the blood–brain barrier to estimate the risk associated with treatment.

There are several new MR techniques that may add new information to the mismatch concept in the future. As for perfusion imaging, arterial spin labeling offers a noninvasive and contrast agent-independent perfusion measure that allows repetitive CBF measurements (Chalela et al, 2000; Viallon et al, 2010). As for MR-based metabolic imaging, measures of oxygen metabolism as well as for tissue pH might serve as a direct estimate of the penumbra but these techniques await further evaluation (Dani et al, 2010; Geisler et al, 2006; Holmes et al, 2011; Sun et al, 2010).

Conclusion

The mismatch concept is a simplification of the concept of core and penumbra but does not reach its pathophysiological specificity. From a methodological point of view, a complete congruence between the methods cannot be expected. However, stroke MRI is the best clinical approximation to the tissue at risk and remains the best performing diagnostic tool for clinical research and therapy. The comparison with PET imaging helps to validate and improve stroke MRI in clinical use. Important methodological issues that may cause errors in mismatch assessment have been defined over the past years and respective solutions have been presented. This includes the choice of the best PW map, the application of thresholds and the quantitative approach to PW maps. This knowledge has to be implemented in future trials design with MR-based patient stratification. Considering the many variables in mismatch definition, a standardization is urgently needed to focus further research and to make stroke imaging studies comparable. The current evidence for mismatch guided delayed thrombolysis might be heterogeneous, but it reflects methodological limitations that have largely been overcome. Therefore, there is a challenge to apply the current knowledge of mismatch imaging in an optimized study design for future stroke trials.