Imaging Oxygen Metabolism in Acute Stroke Using MRI
- First Online:
- Cite this article as:
- An, H., Ford, A.L., Vo, K.D. et al. Curr Radiol Rep (2014) 2: 39. doi:10.1007/s40134-013-0039-3
- 590 Downloads
The ability to image the ischemic penumbra during hyperacute stroke promises to identify patients who may benefit from treatment intervention beyond population-defined therapeutic time windows. MR blood oxygenation level-dependent (BOLD) contrast imaging has been explored in ischemic stroke. This review provides an overview of several BOLD-based methods, including susceptibility-weighted imaging, R2, R2*, R2′, R2* under oxygen challenge, MR_OEF and MROMI approaches to assess cerebral oxygen metabolism in ischemic stroke. We will review the underlying pathophysiological basis of the imaging approaches, followed by a brief introduction of BOLD contrast. Finally, we will discuss the applications of the BOLD approaches in patients with ischemic stroke. BOLD-based methods hold promise for imaging tissue oxygenation during acute ischemia. Further technical refinement and validation studies in stroke patients against positron emission tomography measurements are needed.
KeywordsMRI BOLD Oxygen extraction fraction Cerebral blood flow Oxygen metabolism Ischemia
Cerebral oxygen metabolism is essential for generating a steady supply of energy for the normal neuronal activity of the brain. Occlusion of a cerebral artery leads to reduction of oxygen and substrate supply to brain tissue resulting in ischemia and eventual infarction. Emergent evaluation and treatment are required to forestall rapidly evolving ischemic injury . Depending on the depth of ischemia, time to treatment and response to treatment, ischemic tissue can be classified into core, penumbra and oligemia. Severe ischemia in the core rapidly leads to irreversible injury regardless of treatment. The ischemic penumbra maintains sufficient flow to preserve tissue structure but impair neural function , but rapidly evolves to infarction if therapeutic intervention is not instituted in a timely manner . Oligemic tissue demonstrates decreased flow, but not sufficient to result in tissue injury even in the absence of treatment. A pooled analysis of several clinical trials has demonstrated that benefits of treatment decrease, while risks increase, as a function of time .
Thus far, intravenous (IV) tissue plasminogen activator (tPA) remains the only FDA-approved medication for treating acute ischemic stroke patients. Improved outcomes have been demonstrated in patients treated within 3 h  and more recently within 4.5 h of symptom onset  based on population studies. Several additional trials have failed to demonstrate the efficacy of IV tPA beyond 4.5 h [6, 7, 8]. In light of these clinical trials, the current recommended therapeutic time window of IV tPA for ischemic stroke is 4.5 h. This short time window has greatly limited IV tPA treatment to only 3.4–5.2 % of stroke patients [9, 10]. It has been suggested that endovascular approaches, including thrombectomy, may enhance chances for recanalization. However, a recent randomized controlled trial did not show a treatment benefit in patients who had endovascular embolectomy using MR imaging criteria for treatment . The use of an early generation clot retrieval device might, in part, be responsible for the negative trial results. More recently, the new SOLITAIRE™ Flow Restoration device achieved substantially better vessel recanalization, safety and clinical outcomes compared to earlier generation retrieval devices .
It is likely that the therapeutic window for tPA, endovascular thrombectomy or other potential therapies may vary between individuals , depending on a host of factors such as vascular anatomy, collateral flow patterns, comorbidities and body temperature . Imaging the ischemic penumbra during hyperacute stroke may be one approach to individualize therapeutic opportunities beyond population-defined time windows.
Over the past 2 decades, the magnetic resonance (MR) diffusion-perfusion (DWI/PWI) mismatch (DPM) has been widely used as a surrogate imaging biomarker for the ischemic penumbra [14, 15, 16]. However, it has been demonstrated that the diffusion lesion (the assumed core) may reverse after reperfusion and that DPM overestimates the size of the penumbra [17, 18, 19, 20, 21, 22, 23, 24].
On the other hand, imaging tissue oxygen metabolism may provide a more direct assessment of tissue viability [25, 26, 27, 28] when compared to that offered by DPM. Positron emission tomography (PET) 15O imaging measures the quantitative cerebral blood flow (CBF), oxygen extraction fraction (OEF) and cerebral metabolic rate of oxygen utilization (CMRO2 = CBF × CaO2 × OEF, where CaO2 is the arterial oxygen content). CBF × CaO2 and OEF reflect the oxygen delivery and demand, respectively, whereas CMRO2 reveals the balance between these two. It has been suggested that CMRO2 is a more specific marker for tissue viability [25, 26]. However, while PET 15O imaging is a validated method for measuring regional oxygen metabolism, the 2-min half-life time of 15O and the need for arterial blood sampling has greatly hindered its utilization in the setting of acute stroke. In light of the practical challenge of PET 15O imaging, non-invasive MR blood oxygen level-dependent (BOLD)-based approaches have been explored. In this review, we will introduce the underlying pathophysiological basis of identifying the ‘at-risk’ tissue using oxygen metabolic imaging in acute stroke, followed by MR BOLD contrast. We will review several MR BOLD contrast-based approaches for penumbral imaging in acute stroke.
Hemodynamic and Metabolic Changes During Acute Ischemia
When an artery becomes narrowed or occluded, the mean arterial pressure (MAP) in the distal circulation may fall, depending on the degree of stenosis and the adequacy of collateral blood flow . This decrease in MAP leads to a reduction in cerebral perfusion pressure (CPP) [30, 31]. In order to compensate for decreasing CPP, brain arterioles dilate to reduce vascular resistance in attempts to maintain constant CBF (autoregulation) [32, 33]. As CPP continues to decrease and CBF is maximally increased, a second compensatory mechanism results in an increase in the fraction of total oxygen extracted from blood (OEF) [25, 34] to maintain a close to normal oxygen metabolism. This phenomenon, termed “misery perfusion” , represents ischemic tissue with an elevated OEF and reduced CBF that is “at risk” for future infarction if left treated . Misery perfusion, regarded as a signature of the ischemic penumbra [36, 37, 38], has been identified in some stroke patients up to days after symptom onset : evidence of misery perfusion was found in 100, 83, 57 and 35–45 % of patients within 9, 12, 24 h and 3–4 days after stroke onset . Collectively, these studies support the notion that penumbra may exist in some patients well beyond currently defined therapeutic time windows, but the proportion of patients with viable penumbra diminishes over time at different rates. Therefore, a penumbral imaging biomarker based on oxygen metabolism holds promise for identifying patients who may be eligible for an extended therapeutic time window with a goal of maximizing the treatment benefit and minimizing the associated risk.
Blood Oxygen Level-Dependent Contrast (BOLD)
Blood is composed of plasma, red blood cells (RBCs), various leukocytes (white blood cells) and platelets. Circulating RBCs contain hemoglobin, which is found in both the oxy- (Hb) and deoxyhemoglobin (dHb) states, depending on the binding of oxygen to the heme moiety within the hemoglobin protein complex. Iron in the heme moiety of deoxyhemoglobin is in a high spin ferrous Fe2+ state with four of its six outer electrons being unpaired. The unpaired electron spins induce large magnetic moments that make deoxyhemoglobin an endogenous paramagnetic contrast agent . In contrast, iron in oxygenated hemoglobin has an electron that is partially transferred to the oxygen molecule, resulting in a low-spin state heme iron. Hence, oxyhemoglobin exhibits diamagnetic properties.
The susceptibility difference between deoxyhemoglobin and oxyhemoglobin provides a powerful endogenous MR contrast that is dependent on the blood oxygenation level. Relaxation rates R2 (1/T2), R2*(1/T2*) and R2′ (=R2*−R2) are increased with increasing deoxyhemoglobin concentration [42, 43, 44, 45]. Several MR BOLD methods directly or indirectly measure OEF or oxygen metabolism in an effort to image viable but at-risk tissue.
Susceptibility Weighted Imaging (SWI) in Acute Stroke
R2 or T2 BOLD Contrast
The oxygenation dependence of R2 originates from the irreversible dephasing caused by water exchange and diffusion through the magnetic field gradient induced by deoxyhemoglobin. Using a Carr-Purcell-Meiboom-Gill (CPMG) sequence, Thulborn et al.  demonstrated that R2 (1/T2) of the blood increased along with the increase of the concentration of dHb. Kavec et al. have demonstrated that a drop in CBF from 58 ± 8 to 17 ± 3 ml/100 g/min caused a shortening of T2 from 66.9 ± 0.4 to 64.6 ± 0.5 ms, without diffusion abnormality. The decreased T2 (increased R2) suggested an elevated OEF during misery perfusion. In contrast, a moderately reduced CBF of 42 ± 7 ml/100 g/min did not induce a detectable T2 change . Kettunen et al.  reported that a CBF below 20 ml/100 g/min led to a decrease of T2 of 4.6 ± 1.2 and 6.8 ± 1.7 ms after 7 and 15 min of ischemia, respectively, in a transient common carotid artery occlusion rat model. Other studies have demonstrated that T2 changes are reflected in a U-shaped curve [1, 51•, 52] as a function of CBF. Only moderately reduced CBF leads to decreased T2, while T2 increased in regions with severely reduced CBF [51•, 52]. Based on the pathophysiology of misery perfusion, a shortened T2 is expected but the U-shaped T2 change is unexpected. A plausible explanation of the observed T2 elevation in severe ischemia is vasogenic edema, which might occur very rapidly. Furthermore, the initial drop of T2 was followed by an increase in T2 after 30–60 min of stroke . Vasogenic edema might also be the primary reason for the subsequently increased T2 following the initial reduced T2.
In summary, these studies suggested that the shortened T2 (or increased R2) shortly after ischemia may be caused by an increase of the local concentration of dHb, which is consistent with the predicted physiology of misery perfusion. However, since T2 changes may depend on the brain oxygenation level as well as other pathophysiological processes such as vasogenic edema or inflammation, interpretation of the T2 changes in acute stroke may be very challenging. Moreover, since R2 is not as sensitive as R2* to the change of dHb concentration [54, 55], it is difficult to detect subtle cerebral oxygenation differences using R2-based methods.
R2* or T2* BOLD Contrast
In a simple way, R2* can be depicted as a summation of R2 and R2′ (R2* = R2+R2′), where R2 (1/T2) and R2′ (1/T2′) are the irreversible and reversible relaxation rates, respectively. Signal decay caused by the R2′ relaxation can be recovered in a spin echo. It has been demonstrated that R2* is more sensitive to the deoxyhemoglobin concentration change than R2 [54, 55]. Using a gradient echo (R2*) imaging approach, De Cresipgny et al.  have demonstrated a rapid signal drop upon MCAO occlusion and an overshoot after successful reperfusion in the cat brain. These findings suggest an initial increase followed by a decrease of dHb during acute ischemia and after reperfusion, respectively. In a rat focal ischemia model, areas of low T2* signal were larger than the regions with decreased diffusion , suggesting that regions with low T2* may encompass penumbral tissue. Furthermore, Tamura et al. used signals from dynamic susceptibility contrast imaging (DSC) before contrast arrival to demonstrate that the mean T2*-weighted signal is significantly lower in the affected hemisphere in acute stroke patients [58•]. Similar MRI findings were also reported in two other patient studies [59, 60]. It is postulated that the hypointense area may correspond to regions with elevated OEF under misery perfusion. However, a recent MR PET study [61•] found no significant relationship between the PET-measured OEF and MR T2*-weighted signal obtained from the pre-contrast arrival DSC signal. There are two plausible explanations for this discrepancy. First, background field inhomogeneities can cause an overestimation of R2* in focal regions that is not related to misery perfusion. Second, similar to R2, R2* (R2* = R2 + R2′) can be affected by vasogenic edema.
R2′ BOLD Contrast
R2 and R2* approaches are not specific and could be affected by numerous factors, including tissue oxygenation, brain edema or inflammation. In contrast, R2′ is more directly related to the concentration of dHb and the volume fraction of dHb . Therefore, it may be a better imaging biomarker for measuring OEF during acute ischemia. Geisler et al. examined R2′ in regions with apparent diffusion coefficient (ADC) lesion growth (recruited in the final lesion but without acute ADC abnormality) and surviving tissue [63••]. They observed an elevated R2′ in the ischemic hemisphere with the highest R2′ in the acute ADC lesion, an increased but lower R2′ in the region of lesion growth and the lowest R2′ in surviving tissue regions. More recently, Siemonsen et al. reported that R2′ > ADC mismatch can better predict infarct growth when compared to TTP > ADC mismatch in patients treated with thrombolytics [64••]. Using a 1-h MCA occlusion monkey model, Zhang et al.  evaluated the temporal change of R2′ in the ischemic core (hyperintense signal in both acute DWI and 48-h T2w), penumbra (hyperintense signal in acute DWI without a 48-h T2w lesion) and oligemia (prolonged acute MTT with negative acute DWI and 48-h T2w) during MCAO and after reperfusion. Elevated R2′ was detected in all three regions during MCAO, suggesting a local dHb increase. After reperfusion, R2′ decreased in the ischemic core, while it increased in both the penumbra and oligemia over time . The persistently increased R2′ in the penumbra and oligemia after reperfusion is difficult to interpret. Of note, the definition of core/penumbra/oligemia in this study differs from the conventional definition of penumbra and oligemia in most of the DPM literature.
In summary, results based on the R2′ methods are encouraging and may suggest clinical utility in acute stroke. However, since R2′ may increase in all ischemic regions including the core, penumbra and oligemia (definitions varied among different studies), it remains unclear whether R2′ alone will be useful in distinguishing an ischemic penumbra from core and oligemia. Similar to R2*, R2′ is affected by background field inhomogeneities. Careful system shimming and background susceptibility corrections are needed for an accurate representation of the dHb concentration.
R2* Change After Oxygen Challenge
Given the challenges associated with absolute measurements of R2* or R2′, relative R2* changes before and after oxygen challenge (OC) have been proposed to distinguish ischemic core, penumbra and oligemia. OC is achieved by inhalation of 100 % oxygen under normobaric conditions. The use of ΔR2* or ΔS (S is signal) before and after OC minimizes the effect of background field inhomogeneities. During OC, molecular oxygen may convert deoxyhemoglobin to oxyhemoglobin, therefore resulting in a signal increase in T2*-weighted images (or decrease of ΔR2*). The hypothesis is that a high concentration of dHb in the penumbral tissue (suggestive of high OEF during misery perfusion) may lead to greater signal change in the penumbra. Santosh et al.  reported signal increases of 1.8, 3.7 and 0.24 % in the contralateral cortex, region of DPM and ADC-defined ischemic core, respectively, using a permanent MCA occlusion (pMCAO) rat model. In another study, [14C]2- deoxyglucose (2-DG) autoradiography was used in combination with T2* OC MR to determine the glucose metabolic status of various regions . The ischemic core showed low T2* OC signal change and low glucose utilization as demonstrated in 2-DG images. In regions with increased T2* signal during OC, a normal glucose metabolism similar to that of the contralateral hemisphere was found. In addition, a hyperglycolytic band was detected within the ADC lesion. A small T2* OC signal change lower than that from the contralateral hemisphere was obtained in this hyperglycolytic band. Similar T2* signal changes resulting from OC were found in rats following pMCAO . Using a transient MCA occlusion (tMCAO) rat model, Robertson et al.  examined the T2* signal response to OC during ischemia, immediately and 7 days after reperfusion. In this study, the ischemic core and penumbra were defined in regions with low (core) or high (penumbra) T2* OC responses during ischemia, respectively. They found that the T2* OC response remained low in the core before and after reperfusion. The ischemic core infarcted 7 days after reperfusion. The contralateral hemisphere had consistently moderate T2* signal change during ischemia and after reperfusion. In contrast, penumbra had an elevated T2* OC response during ischemia but then a decreased OC response upon reperfusion. The OC-defined penumbra (high OC response during ischemia) was not recruited into the final infarct. A more recent study reported a different OC response evolution in core regions (defined in regions with low OC response during ischemia) after reperfusion in rats undergoing a tMCAO . In this study, the core regions had low OC during ischemia, but high OC responses after reperfusion. Since most of the core regions became infarcted, this suggests that high OC responses, if used alone, do not necessarily reflect salvageable tissue.
Dani et al. [71••] have reported the first clinical stroke application of the OC approach. Thirty-five stroke patients with a median onset time of 18 h were recruited. Data from 18 patients who had reasonable T2* signal time curves in the unaffected hemishpere were included in the analysis. T2* OC signal curves were evaluated in the presumed ischemic core (defined by abnormal ADC), ischemic penumbra (perfusion–diffusion mismatch or expanded ADC lesion at day 3) and contralateral hemisphere. In general, the DWI-defined core showed a reduced OC response as compared to contralateral tissue, while the presumed penumbra in patients who were imaged acutely (<8 h) showed a trend of greater OC responses. However, a large variation in OC responses was observed across patients. Ten, 4 and 3 out of 17 DWI core lesions showed positive, flat and negative OC responses, while 5, 1 and 2 out of 8 penumbra lesions showed positive, flat and negative OC responses, respectively. Moreover, there was poor spatial correspondence between the DPM-defined penumbra and regions with high OC responses in most of the patient examples shown in [71••]. The subpar performance of this OC approach in clinical stroke is not surprising, given that stroke patients have much greater variations in hemodynamics and the elapsed times after stroke onset compared to the experimental MCAO in animal models.
In summary, encouraging results have been reported that potentially distinguish core, penumbra and normal tissues using the T2* signal change to OC, particularly in experimental MCAO. Since several factors including CBV, CBF, reperfusion status, reperfusion time and T1 change (due to an elevated level of dissolved O2 in blood during OC) may all affect the OC responses in human stroke; therefore, it is challenging to use simple OC response thresholds to separate penumbra from core, oligemia and normal tissues without considering these other factors. The advantages of this OC approach include (1) the reduced confounding influence of background susceptibility and (2) wide availability in most clinical centers by using a standard T2* MR pulse sequence.
MR_OEF and MR_OMI
All the aforementioned methods using the SWI, R2*, R2, R2′ or R2* with OC are indirect measurements of tissue oxygenation, which may be subject to other factors that are irrelevant to tissue metabolic activities. On the other hand, as suggested by the PET literature, OEF and CMRO2 can provide a more direct assessment of tissue viability and therefore may be better imaging biomarkers for penumbral imaging.
Extensive efforts have focused on measuring brain OEF and CMRO2 using MR approaches at the whole brain (global) level or at the regional level. Since whole brain global measurements have little utility in focal ischemia, we will only review MR approaches for measuring regional OEF that have shown promises in stroke. As suggested by the quantitative BOLD (qBOLD) biophysical signal models [72, 73], R2′ is proportional to the product of the concentration of dHb and the venous cerebral blood volume (vCBV). Regional cerebral venous oxygen saturation (SvO2) or OEF (SaO2-SvO2) can be measured if both R2′ and vCBV can be acquired. Using either a gradient echo sampled spin echo [73, 74] or an asymmetric spin echo EPI sequences (ASE) , PET comparable OEF was obtained from normal subjects under both normal and hypercapnic conditions [74, 75, 76]. MR-measured cerebral oxygen saturation has been validated against gold standard blood gas analysis for a wide range of oxygen saturations (0.1–0.8) in rats under moderate, severe hypoxic, hypercapnic and normal control states [77••]. Similarly, good correlation between the MR measured oxygen saturation and blood gas oximetry was obtained in an SvO2 range of 0.45–0.85 induced by two different anesthetic conditions using isoflurane and alpha-chloralose anesthesia . Moreover, using multiple MR scans to estimate R2*, R2 and CBV separately, Christen et al.  demonstrated that MR-measured SvO2 correlated well with blood gas oximetry results in rats under hyperoxia and hypoxia. Of note, the arterial-venous oxygen difference was also the same gold standard by which PET-OEF was validated [80, 81].
BOLD-based MR_OEF methods measure susceptibility, a parameter related to the concentration of dHb, and infer oxygen extraction fraction from this measurement assuming that SaO2 in large arteries and arterioles is at 100 % even during focal ischemia. Since both OEF and vCBV affect the MR signal, an accurate measurement of vCBV is needed in order to obtain OEF. A high signal-to-noise ratio (SNR) is usually required [82, 83]. MR_OEF has been measured in healthy human subjects under both normal and hypercapnic conditions, and values are closely comparable to those reported using PET . Moreover, global validation of MR_OEF has been performed in animals [78, 79, 84]. However, regional MR_OEF measurement has not been validated against simultaneous PET OEF under pathophysiological conditions.
MR-derived oxygen metabolism can be computed as a product of MR_OEF and CBF. This parameter was first termed as MR_CMRO2  to distinguish it from the PET CMRO2 (defined as OEF × CBF × CaO2). The ‘MR-derived cerebral oxygen metabolic index’ (MR_COMI or simply MROMI) has since been utilized [77••].
CBF can be obtained using either DSC  or arterial spin labeling (ASL) [86, 87, 88] methods. The DSC approach requires the injection of a contrast agent, while the ASL approach uses magnetically labeled arterial blood to derive CBF. The DSC method is widely utilized in clinical stroke because of its speed and good SNR. However, absolute CBF measurement based on DSC is very challenging. In contrast, ASL methods provide absolute CBF measurement but with a much lower SNR.
MROMI has been used to depict temporal evolution of cerebral oxygen metabolism during cerebral ischemia in rats subject to tMCAO [77••]. In this focal ischemia study, MROMI values were lowest within the infarct, while the surviving peri-infarct region showed moderately reduced MROMI. MROMI within the region of ischemia evolved in space and time. Ischemic regions with severely reduced MROMI (presumed core) continued to expand while regions with moderately reduced MROMI (presumed penumbra) decreased in size over time. These observations suggested that the metabolically active penumbra evolves into the infracted core as time progresses.
The potential clinical utility of MROMI was first evaluated in seven acute stroke patients . Normalized MROMI values were 0.40 ± 0.24 and 0.55 ± 0.11 of the contralateral hemisphere in the final infarct and the DPM regions, respectively, in agreement with previous PET measurements in stroke patients .
To test the clinical utility of MROMI, a prospective imaging study was undertaken. Forty acute stroke patients were serially imaged 3.0 h (tp1), 6.2 h (tp2) and 1 month (tp3) after symptom onset [90••]. CBF was determined using DSC and OEF was obtained using the ASE method; MROMI was computed as the product of CBF and OEF. In the tp1 MRI (obtained within 3 h of stroke onset) regions with very low MROMI values predicted eventual infarction of tissue as demonstrated on the 1-month FLAIR image. To further investigate whether MROMI can be utilized to delineate the ischemic penumbra, a quantitative analysis was designed to find threshold values that optimally segregated tissue into core, penumbra and oligemia, based on ideal infarct probabilities. In theory, the infarct probability (IP) for tissue in the core is 100 %; for non-reperfused penumbra, 100 %; for reperfused penumbra, 0 %; for oligemic tissue, 0 %. Average prediction error (APE, averaged difference between the actual IP and ideal IP across the four tissue groups) was utilized to choose an optimal pair of MROMI thresholds to define core, penumbra and oligemia [91••]. A pair of optimal DPM thresholds was also derived through a similar process so that DPM thresholds could be compared directly to OMI thresholds for their ability to predict reperfusion-dependent infarction. We found that the MROMI outperformed DPM in predicting reperfusion-dependent infarction (minimizing APE) in this cohort of patients [92••].
In summary, preliminary results supporting the use of MROMI for penumbral imaging are encouraging. MROMI requires the measurement of both OEF and CBF, in addition to tissue segmentation, which is required for normalization to minimize its dependence on tissue type. In addition, a custom sequence was utilized for the measurement of OEF. The complex imaging acquisition and data analysis have made the method sensitive to noise and imaging artifacts. A robust, rapid and streamlined data acquisition and post-processing method will improve the clinical utility of this technique. Furthermore, validation of regional measurement of MROEF and MROMI by direct comparison against PET measurement is needed, especially in patients with focal lesions. Since hemodynamic changes occur rapidly during hyperacute stroke, an integrated MR/PET system that allows simultaneous acquisition will be valuable for this validation.
Noninvasive approaches providing information on tissue viability during acute ischemic stroke have been extensively pursued. In this review, we introduce several MR BOLD imaging approaches to directly or indirectly measure tissue oxygenation and oxygen metabolism. While results are encouraging, further technical refinement and validation in patients are warranted to fully test whether these techniques can be translated for clinical use in acute ischemic stroke.
This study was supported in part by grants from the National Institute of Health (NIH 5R01NS054079, NIH 5P50NS055977) and American Heart Association (AHA 0730321N).