Clinical Correlates of Proton Magnetic Resonance Spectroscopy Findings After Acute Cerebral Infarction
Background and Purpose We sought to determine whether lactate and N-acetyl signals measured by proton magnetic resonance spectroscopy (MRS) in the first days after stroke correlate with clinical measures of disability and functional outcome.
Methods One-dimensional spectroscopic imaging was performed after stroke on 32 patients using a 2.1-T magnet. The Toronto Stroke Scale score at the time of the MRS study and the Barthel Index score at hospital discharge were determined from patient records. Lesion volume was estimated by a tracing algorithm from the scout magnetic resonance image obtained as part of the MRS study. The scaled lactate and N-acetyl signals from the voxel having the highest measured lactate were used to predict the clinical variables and lesion volume, as well as relative perfusion within the lesion, in those patients who underwent single-photon emission computed tomography (SPECT) blood flow imaging, using a multiple regression analysis. The correlation of lesion volume with the clinical variables was also evaluated.
Results Lesion lactate signal was correlated with the Toronto Stroke Scale score, Barthel Index score, lesion volume, and SPECT score, all at P<.01. The N-acetyl level correlated with the Barthel Index score and lesion volume at P<.05. Lesion volume was also strongly correlated with the clinical variables (P<.0001).
Conclusions This is the first study to document the clinical predictive value of proton MRS measurements in patients after stroke. The association with functional outcome is stronger for lactate than for N-acetyl. Spectroscopic assessment of the metabolic status of cerebral tissues shortly after infarction may have significant clinical utility.
Stroke management is most effective when guided by early, accurate diagnosis, a fact that will rise in importance as better therapies are introduced. Magnetic resonance spectroscopy (MRS) can supply chemically specific information about an acute infarct as soon as a patient with stroke syndrome can be admitted to a hospital. Combined with magnetic resonance imaging (MRI) in the same session on the same machine, such data ought to have considerable prognostic utility, but so far neither 31P nor 1H studies1 2 have demonstrated that value. We report here on the correlation between the lactate and N-acetyl (NA) signals from initial 1H MRS examination and other measures of lesion severity, most notably with functional outcome.
Subjects and Methods
Imaging and Spectroscopy
All studies were performed on a 1-m-bore modified Bruker-ORS Biospec 1 spectrometer operating at 2.1 T and fitted with a birdcage head resonator. An eight-image axial T2-weighted MRI (slice thickness, 5 mm; interslice separation, 2 mm; repeat time [TR], 2.05 seconds; echo time [TE], 95 milliseconds; 128×128 acquisition matrix; field of view, 19.5×19.5 cm2) was used to identify the location of the stroke and for lesion volume estimation. This limited imaging protocol was chosen to minimize total examination time, most of which was required for spectroscopy. The stimulated-echo acquisition mode (STEAM) pulse sequence3 (TE, 270 milliseconds; TR, 4 seconds; mixing time [TM], 70 milliseconds; 256 iterations) was then applied over a 3×6×3-cm3 or, in later studies, a 3×8×2-cm3 region of interest. The long TE of 270 milliseconds was chosen to optimize the J-modulating lactate resonance and to eliminate any signals from lipids. An inversion recovery pulse 800 milliseconds before the start of the STEAM sequence, and a frequency selective three-lobe sinc pulse and crusher gradients applied during TM, were used to achieve water suppression. Phase encoding over the long axis (perpendicular to the sagittal plane) generated a one-dimensional spectroscopic image (SI) of 10 voxels, each with a nominal volume of 5.4 or 4.8 cm3. The average postshimming spectral line width was 5 to 6 Hz per voxel. Total examination time (including the scout image, SI region selection, localized shimming, and SI acquisition) averaged about 45 minutes. Fig 1⇓ shows a representative MRI from patient 19, obtained 18 hours after onset of weakness on the left side, numbness, and dysarthria. The SI region of interest is superimposed on the MRI; each rectangle in the figure represents a single voxel in cross section. The voxel with maximum lactate used in the statistical analysis is highlighted, and the spectrum obtained from this volume is shown in Fig 2⇓.
The signal was acquired as a full echo and was sine-bell apodized in both dimensions to improve spectral signal-to-noise before Fourier transformation. A magnitude calculation was then performed on the SI series to obviate the need for individual phase correction of each spectrum in the frequency domain. Signals were quantified by numerical integration of the area under each peak and scaled to NA in the contralateral normal hemisphere.4 Observed peaks were identified by chemical shift and, in the case of lactate, by the characteristic doublet with a splitting of 7 Hz when spectra were processed by exponential multiplication of the free induction decay signal before Fourier transformation to enhance spectral peak resolution.5
The volume of the stroke was estimated by summing the number of pixels within the area of signal abnormality as seen on each T2-weighted scout MRI slice, using a computerized, user-driven tracing routine. The pixel count in each axial section was converted to an area given a pixel size of 2.3 mm2, and the areas from all of the slices containing the ischemic tissue were combined to generate a volume estimate based on a 7-mm slice separation.6 No attempt was made to correct the estimate for lesions that extended beyond the scout image.
A total of 32 patients were included in the analysis. Of these, 4 had a lacunar infarct, and all others had cortical infarcts. Patients with primary intracerebral hemorrhage as assessed by computed tomography (CT) or MRI were excluded from the study. All patients had a new motor or speech deficit and evidence of acute stroke on CT or MRI and were first examined an average of 4.9 days after symptom onset (range, 0 to 19 days). The average age was 62.6 years (range, 45 to 82 years). Some of these MRS data, analyzed in a different manner, have been reported in two prior publications.4 6 Informed consent was obtained from all study participants or from the closest family member if the patient was unable to give informed consent because of neurological deficits. Our research protocol was reviewed and approved by the Human Investigation Committee of the Yale University School of Medicine.
The Toronto Stroke Scale (TSS) score,7 a quantitative measure of current neurological deficit, was calculated from hospital records for 31 subjects at the time of the MRS examination. The TSS evaluates 16 components of the neurological examination, with possible scores ranging from a total of 0 (normal for all items) to 44 (maximum impairment). The Barthel Index score,7 which assesses functional outcome after stroke, was determined at time of hospital discharge in all but 1 of the patients. The Barthel Index rates a patient’s ability to perform 10 activities of daily living on a scale of 0 (unable to perform any of the tasks) to 100 (completely independent for all tasks). The clinical scale scores were determined by a single neurologist reviewer from detailed hospital medical records using a standard set of evaluation criteria. Sixteen of the subjects also had cerebral blood flow scans with 99mTc-hexamethylpropyleneamine oxime (HMPAO) single-photon emission computed tomography (SPECT) within an average of 2.8 days of their MRS study (range, 0 to 8 days). The SPECT results were graded on a qualitative scale from 1 to 4, where 1 represents a finding of hyperemia around the lesion and 2 through 4 signify slight, moderate, or severe hypoperfusion, respectively.
The scaled lactate and NA signals in the SI voxel located within the lesion with the largest lactate signal were the MRS-derived parameters used in the analysis. Other variables examined were patient age, the day of MRS study, and the day of SPECT examination after stroke symptom onset. Only the spectroscopic results and TSS score at the time of initial study were used for patients examined by MRS more than once. These values were used to predict independently the TSS score, Barthel Index score, lesion volume, and SPECT score from a multiple regression analysis. Lesion volume was also used to predict the TSS and Barthel Index scores in a separate least-squares computation. Differences were considered statistically significant at P≤.05. The raw data values used in the analysis are listed in Table 1⇓.
As shown in Table 2⇓, lesion lactate was highly correlated with both acute stroke severity (as assessed by lesion volume, TSS, and SPECT perfusion deficit) and eventual clinical outcome (as measured by the Barthel Index at the time of hospital discharge). Lesion NA was not correlated with the TSS or SPECT scores but was correlated with lesion volume at a lower level of significance than lactate. NA was most significantly correlated with stroke functional outcome, reflected in the Barthel Index score. Plots of the Barthel Index score versus the MRS-measured lactate and NA values are shown in Fig 3A⇓ and 3B⇓, respectively. The line in each graph is the projection of the fitted multiple regression line into the two-dimensional space of the plot. Other variables, such as patient age and day of MRS examination, were not significantly correlated with the independent variables examined when added to lactate and NA in the multiple regression analysis. In those patients for whom both internal and external concentration standard spectra were available for scaling, lactate values scaled using the two methods were very closely correlated (r=.983), verifying the consistency of the internal contralateral hemisphere NA scaling used to generate the MRS data for our analysis.
Lesion volume was also found to be highly correlated with scores of both TSS (r=.666, P<.0001) and discharge Barthel Index (r=.734, P<.00001), based on linear least-squares analyses. These correlations are illustrated in Fig 4⇓.
Lactate and NA measurements by 1H MRS correlate with other initial measures of lesion severity and appear to predict functional outcome. The stronger correlations are with lactate increase rather than NA decline. This difference is consistent with our observation that the maximum lactate concentrations associated with a stroke are present immediately after stroke onset, whereas the NA signal falls from its normal high value over a matter of days,4 6 presumably as debris from destroyed neurons is cleared. The lactate signal, therefore, might be expected to provide the earliest MRS-measurable indicator of ultimate lesion severity. Since high concentrations of lactate are produced during the period of initial ischemic injury,8 before significant lesion infiltration by leukocytes,9 10 its level and anatomic extent might be expected to correlate with both the degree of acute neurological impairment and degree of permanent cerebral injury, reflected in functional outcome.
Partial volume effects may also contribute to the correlation between MRS and clinical variables, since more extensive lesions will more completely fill the nominal 4.8 to 5.4 cm3 volume of the SI voxel and so influence determination of the highest lactate. Spectroscopic studies of small lesions will incorporate larger amounts of normal brain, which has no measurable lactate signal. Cerebral N-acetylaspartate, the principal contributor to the in vivo NA signal, is located principally if not exclusively within neurons11 12 and is depleted from brain lesions13 14 ; the residual NA signal seen on MRS more than several days after stroke often arises from surrounding normal brain rather than from the infarct itself.15 Thus, a decreased NA signal on an initial study suggests that neurons have already been lost, leaving functional recovery dependent on surviving neurons. With the finer anatomic resolution that will become available as SI technology evolves, early NA changes that are relatively free of partial volume effects are likely to achieve important prognostic utility.
Few studies have explicitly compared poststroke MRS findings with clinical measures. Gideon et al2 found no clear relation between acute lactate and NA levels within a stroke lesion and clinical outcome.2 However, their study was based on a total of six patients, half of whom had infarcts of modest to small size, and they did not use quantitative assessment of clinical status. Because neurological disability and clinical outcome vary considerably among individual patients even with lesions of similar size, depending on stroke location and other factors, it is reasonable to expect that a larger sample size would be needed to demonstrate a statistically significant correlation at the present relatively coarse anatomic resolution of MRS. Perhaps for similar reasons, a 31P MRS of 21 patients after stroke found trends toward a positive correlation between the ratio of inorganic to total phosphate and a measure of current neurological deficit (the Hemispheric Stroke Scale16 ) and an inverse correlation with the Barthel Index, neither of which, however, attained statistical significance.1
This study is the first of which we are aware to document a correlation between proton MRS findings and clinical variables. However, several limitations should be kept in mind. The standardized combination of data from many patients is necessary to attain statistical power but involves a simplification of the individual differences between patients. Early T2-weighted MRI after stroke reflects local edema as well as ischemic and infarcted tissue and therefore may overestimate the size of the ultimate lesion. The relatively large voxel size used and the limitation of our analysis to a single SI volume did not permit distinctions between maximum lactate located in core versus peripheral regions within the stroke. In our retrospective analysis, the MRS examination and the SPECT study (when done) were performed over a range of times after the stroke, making comparisons between studies and between patients more difficult. Patients studied initially by MRS some days after stroke may already have lower lactate levels, since lactate tends to decrease with time.4 6 This additional source of variability inherent in a retrospective review suggests that our results may underestimate the true correlation between the MRS and clinical measures. Although the estimation of TSS and Barthel Index scores from hospital records was performed in a self-consistent manner from detailed clinical data, the discharge-day Barthel Index score used in our analysis involved a short follow-up period and was obtained on a different day for each patient. A standardized follow-up evaluation at 3 months, or even 1 month, after stroke would provide a more reliable and uniform measure of ultimate recovery. In general, however, those patients having early hospital discharge, and therefore the briefest follow-up period to estimation of the Barthel Index score, had the mildest deficits and the greatest degree of functional recovery, often achieving a Barthel Index score of 100. A longer follow-up period in these patients would be unlikely to change their score. The validation of our results through a prospective study assessing the correlation between MRS-measured lactate, NA, and perhaps other metabolites and clinical variables is desirable.
Lesion volume also correlated well with the measures of disability and clinical outcome that we examined. As clinical trials begin to use MRI measures as intermediary end points in the evaluation of new treatments for stroke, documented correlations between lesion size and clinical deficit or functional outcome should help to predict the magnitude of decrease in lesion size required to produce a clinically significant improvement. Diffusion-weighted imaging may permit visualization of infarcts and computation of lesion volumes within minutes of stroke,17 but its cellular correlates and hence its clinical significance remain to be determined. In vivo diffusion measurements are also highly sensitive to subject motion.18 Definition of lesion boundaries, whether done manually or by computer algorithm, tends to be somewhat arbitrary and subjective.
In contrast, early spectroscopy provides, in a single measurement of lactate, data that are of plausibly interpretable pathophysiological significance and that portend much for the clinical status of the patient. Whereas the MRS values reported here were extracted from one-dimensional spectroscopic images, data from a single volume centered within the stroke were used for our analysis. Single-volume spectroscopy is technically easier to perform than one- or two-dimensional SI, and highly automated algorithms for acquiring single spectra on standard clinical MRI machines have been developed that could be added to an initial clinical MRI study after stroke. An investment of modest additional examination time might yield important early prognostic data to help guide patient management. Conversely, SI in two dimensions would define the spatial extent of metabolic derangements surrounding an ischemic lesion. In conjunction with other MRI measures, particularly diffusion-weighted imaging, SI should confer unprecedented predictive accuracy to the early evaluation of patients with cerebral infarction. The possibility certainly deserves prospective study.
This study was supported by National Institutes of Health grants NS-21708 and DK-34576, National Institutes of Health Fogarty International Center grant 1F05TW04640 (Dr Kalvach), and a National Stroke Association fellowship award (Dr Blamire).
- Received August 3, 1994.
- Revision received November 16, 1994.
- Accepted November 17, 1994.
- Copyright © 1995 by American Heart Association
Gideon P, Sperling B, Arlien-Soborg P, Olsen TS, Henriksen O. Long-term follow-up of cerebral infarction patients with proton magnetic resonance spectroscopy. Stroke. 1994;25:967-973.
Frahm J, Merboldt K-D, Hanicke W. Localized proton spectroscopy using stimulated echoes. J Magn Reson. 1987;72:502-508.
Graham GD, Blamire AM, Howseman AM, Rothman DL, Fayad PB, Brass LM, Petroff OAC, Shulman RG, Prichard JW. Proton magnetic resonance spectroscopy of cerebral lactate and other metabolites in stroke patients. Stroke. 1992;23:333-340.
Harris RK. Nuclear Magnetic Resonance Spectroscopy. New York, NY: John Wiley & Sons, Inc; 1986.
Graham GD, Blamire AM, Rothman DL, Brass LM, Fayad PB, Petroff OAC, Prichard JW. Early temporal variation of cerebral metabolites after human stroke: a proton magnetic resonance spectroscopy study. Stroke. 1993;24:1891-1896.
Wade DT. Measurement in Neurological Rehabilitation. 1st ed. New York, NY: Oxford University Press; 1992.
Felber SR, Aichner FT, Sauter R, Gerstenbrand F. Combined magnetic resonance imaging and proton magnetic resonance spectroscopy of patients with acute stroke. Stroke. 1992;23:1106-1110.
Gideon P, Henriksen O, Sperling B, Christiansen P, Olsen TS, Jorgensen HS, Arlien-Soborg P. Early time course of N-acetylaspartate, creatine and phosphocreatine, and compounds containing choline in the brain after acute stroke: a proton magnetic resonance spectroscopy study. Stroke. 1992;23:1566-1572.
Adams RJ, Meador KJ, Sethi KD, Grotta JC, Thomson DS. Graded neurologic scale for use in acute hemispheric stroke treatment protocols. Stroke. 1987;18:665-669.
Warach S, Chien D, Li W, Ronthal M, Edelman RR. Fast magnetic resonance diffusion-weighted imaging of acute human stroke. Neurology. 1992;42:1717-1723.