Magnetic Resonance Versus Computed Tomographic Imaging in Acute Stroke
Background and Purpose This study was an attempt to determine whether CT and MRI are comparable or if one is superior to the other in the early detection of ischemic stroke or hematoma.
Methods Patients with acute stroke were sought within 3 hours of onset for clinical examination and prospective evaluation by concurrently performed CT and MRI. Repeated clinical and imaging studies were undertaken when possible immediately after imaging and at 24 hours, 3 to 5 days, and 3 months. The study neurologists were blinded to the results of imaging, as were the study radiologists to the clinical findings. The study radiologists read the scans in sequence, mapping each imaging on standard templates before viewing a later scan. No retrospective revisions of imaging mapping of earlier images were undertaken.
Results Sixty-eight patients were recruited within 4 hours and an additional 12 patients within 24 hours. Seventy-five strokes were due to infarction and five to hemorrhage. The median time to first scan was 132 minutes. Although some of the infarctions in 75 patients were detected within 1 hour, the fraction of positive first scans approached an asymptote at 2 to 3 hours. Overall, with the use of conventional non–contrast-enhanced CT and T1- and T2-weighted MRI, neither was superior in the very early detection of either hematoma or infarction. There was a marginally significant correlation between early positive brain imaging and the severity of the stroke. Some patients had initially positive CT and/or MRI scans, but their neurological examination had returned to normal by 24 hours. Overall, CT was better than baseline MRI at predicting 24-hour outcome. After 24 hours, both CT and MR more conspicuously defined the lesion limits than they did at baseline.
Conclusions With the technology available through 1991, neither CT nor MRI proved superior in the detection of the earliest signs of stroke.
In the absence of a definitive therapy for acute ischemic stroke, efforts are still under way to learn the timetable of brain injury. Leading among these efforts is an attempt to determine the correlation between clinical evaluation and brain imaging. Although the clinical outcome is the ultimate test of the effectiveness of any treatment modality, an ancillary measure that differentiates changes in lesion size or character on brain imaging would be welcome. For imaging to be helpful, it should ideally measure very early findings of stroke and also be sensitive to early pathophysiological changes.
Animal models have shown that the timetable for infarction is spread over several hours, during which time the core lesion appears and enlarges to its final size.1 The human equivalent of this process has not been well documented by image technology.
Early experience with CT scan was disappointing because studies were initially often negative and only became positive after days.2 With the introduction of contrast enhancement, changes within 2 days were documented.3 With an improvement in technology, CT scans done a few hours from onset in the National Institute of Neurological and Communicative Disorders and Stroke (NINCDS) Stroke Data Bank project showed a remarkably high frequency of findings clinically related to the stroke symptoms. The conclusion from these findings was that the limited capabilities of the early equipment explained the reported low frequency of early positive scans.4 In one study of 36 patients with middle cerebral artery infarction, 70% of CT examinations performed within 4 hours of stroke were positive.5
With respect to MR technology, experimental stroke models in cats and primates that used the T2-weighted technique showed parenchymal signal changes as early as 30 minutes and changes in all cases by 4 hours.6 Recent studies in animal models that used contrast-enhanced and diffusion-weighted techniques demonstrated changes as early as 10 minutes after middle cerebral ligation.7 The increased MR signal images have been correlated with histochemical studies documenting increase in brain water. In these experiments, brain edema occurred and increased before structural changes occurred.8 9 In clinical studies the sequence of signal changes has been similar to that in animal studies, but the time course has been more prolonged.
The present study is our analysis of T2-weighted parenchymal signal change in patients with hemorrhagic and nonhemorrhagic strokes imaged within 6 hours of clinical onset of stroke. Ours appears to be the largest cohort studied this early that has been reported to date.
Subjects and Methods
Patients seen at the two participating institutions (Columbia University and University of Iowa) were eligible for inclusion if they were aged 18 to 90 years and had clinical evidence of a stroke thought attributable to ischemia or intracerebral hemorrhage, occurring outside or inside a hospital setting, and if the time from onset of the new symptoms or signs could be confidently established by a witness to the event to have been less than 180 minutes before the initial hospital evaluation.
We excluded patients with a clinical diagnosis of subarachnoid or traumatic hemorrhage, brain hemorrhage into a known intracranial neoplasm, the presence of other major disease requiring active medical intervention that would prevent compliance with the full protocol, known pregnancy, or an unwillingness to participate in the study.
After an initial brief examination, the patient was transported to the CT and MRI scanners by the clinical investigators and there underwent an examination that used the National Institutes of Health (NIH) Stroke Scale and the supplementary motor examination scale immediately before the patient was admitted to the unit. When possible, the patient underwent the companion test (MRI after CT or vice versa) after completion of the first study. After the second imaging study another complete clinical examination was performed, and the patient was transported to an inpatient unit for further evaluation. There the patient remained under observation, during which time clinical observations were made for evidence of worsening and recurrence. Worsening was defined as an increase in any of the 13 parameters in the NIH Stroke Scale or any of the eight parameters of the muscle strength evaluation of one category or more (eg, Stroke Scale, best gaze category, changed from 1 [partial gaze palsy] to 2 [forced deviation]; item wrist, right, changed from 2 [muscle moves joint against resistance] to 3 [muscle moves joint against gravity]). Recurrence was defined as the appearance of a deficit believed by the examiner to represent either an ischemic or hemorrhagic focus in a vascular territory or anatomic region that differed from the initial syndrome.
Follow-up imaging studies were undertaken 24 hours after the first scan. Thereafter, scanning was performed at 3 to 5 days for those with hemorrhage on the initial CT scan, at 7 to 10 days (with and without contrast) for those with infarction, and again at 3 months for all survivors. Clinical assessments were done in conjunction with the imaging studies.
CT scans were obtained with 5- to 8-mm contiguous sections for the supratentorial compartment and 3- to 5-mm slices for the brain stem and cerebellum, producing a total of 18 images for each series.
Proton MRI was conducted with two T2-weighted sequences, a T1-weighted sequence, and a “balanced” pulse sequence with the use of a high-resolution data acquisition matrix. These methods were used for both infarction and hemorrhage. The T2-weighted series was taken at repetition time (TR) ≥3000 milliseconds and echo time (TE) ≥80 milliseconds, the T1-weighted series at TR ≥700 milliseconds and TE=38 milliseconds, and the balanced series at TR=3000 milliseconds and TE=38 milliseconds with the use of a 0.5-T (University of Iowa) or 1.5-T (Columbia University) field strength. The examinations were performed in the axial plane to produce fifteen 5-mm contiguous sections for each sequence.
At Columbia University the MR work was undertaken with the Phillips 1.5-T prototype used for the development of their commercial devices. At the University of Iowa most of the MR studies were done with a Picker Vista at 0.5 T, but the later study images were conducted with a GE Signa at 1.5 T.
The lesions seen on both image techniques were traced on templates, controlling for head angulation at 5°, 10°, 15°, 20°, and 30° according to a method developed at the University of Iowa10 by investigators blinded to the clinical presentation. In each case the images were plotted in the order they were taken, with no retrospective revisions. The volume determination for MRI used both the T1- and T2-weighted images.
All clinical and imaging data were entered on standard data collection forms, and computed data were sent from the two clinical centers to KAI in Rockville, Md, for data management.
When continuously measured variables (such as time from stroke onset to CT or MRI scan) were not normally distributed, they are described below in terms of medians rather than means. Frequencies of positive and negative scans in relation to other factors were evaluated by means of χ2 tests. Group differences in continuous measures such as the Stroke Scale were examined by means of t tests. Relationships between pairs of continuous variables were evaluated with Pearson’s correlation coefficient.
Description of the Sample
Eighty patients were recruited into the study. The stroke was caused by hemorrhage in 5 patients and by infarction in the remainder. Of the subtypes of infarction, cardiac embolism was the initial diagnosis in 27 patients, large-vessel atheroembolism in 10, large-vessel atherothrombosis in 13, inferred due to small-vessel (lacunar) occlusion in 9, and ischemic stroke of unknown cause in 16. The clinical syndrome was classified as major hemispheric in 35, minor hemispheric in 11, brain stem in 17, deep/lacunar in 15, and was not described in 2.
Patients’ scores on the NIH Stroke Scale at baseline ranged from 1 to 40, with a median score of 7.5. Scores for the additional 8 muscle strength items added to the Stroke Scale ranged from 0 to 40 (median score, 4.0). These scores showed no significant differences between the two research sites (Columbia University and University of Iowa). Thirty patients had had a prior stroke, 11 of whom still showed residual clinical signs.
Time Elapsed From Stroke Onset to Baseline Scans
Patients were imaged first with CT or MRI depending on which test was the most readily available. This resulted in 45 CTs first (10 only, 35 paired with MRI) and 35 MRIs first (4 only, 31 paired with CT). The median time from the onset of stroke to the start of the patient’s first scan (whether CT or MRI) was 132 minutes (range, 1 minute to 9.5 hours); for those patients seen within the planned 180-minute time frame, the median time to first scan was 103 minutes (range, 1 to 180 minutes). The median interval between the start of the first scan and the start of the second scan for all patients was 72 minutes (range, 8 minutes to 11.4 hours). Comparison by sites showed that the first scan was accomplished faster at Iowa than at Columbia (medians, 103 and 143 minutes, respectively), but the interval between the first and second paired scan was shorter at Columbia than at Iowa (medians, 52 and 91 minutes, respectively).
For the 75 patients with infarction, Fig 1⇓ shows the total number of first scans and the number of positive first scans in 1-hour intervals for the first 6 hours after stroke onset. Although lesions were detected within the first hour (1 on CT, 1 on MRI), the fraction of positive first scans was greatest (57%) between 2 and 3 hours after the stroke occurred.
CT Versus MRI at Baseline
For all 5 patients with parenchymatous hemorrhage, the initial CT scans were positive, as expected. For the 61 patients with infarction who successfully underwent both CT and MRI scans at baseline, neither CT nor MRI proved superior (26 positive CT versus 31 positive MRI scans; χ2=0.82; P=NS). Fig 2⇓ shows the percentage of positive CT versus MRI scans for patients with infarction within the first 6 hours after stroke onset. When the infarction group overall was broken down by clinical syndrome (convexity or deep/brain stem) and by initial ischemic stroke diagnosis (as noted above), differences between the proportion of positive first scans on CT versus MRI for infarction were not significant. The number of cases for this comparison was small, which may explain the lack of a statistically significant difference.
Relation of Scan Findings to Stroke Scale and Initial Diagnosis
Patients with at least one positive scan at baseline (CT or MRI) had marginally higher scores on the NIH Stroke Scale than patients with negative scans (mean±SD, 10.3±7.2 versus 7.2±6.5, respectively; t=1.96, P=.054). The motor strength scale (separate from the NIH Stroke Scale) scores, although also tending toward higher levels for patients with positive scans (8.8±8.0) than for those with negative scans (5.9±7.3), did not show significantly different values (t=1.62, P=.11). There was no significant correlation between the time to the first appearance of infarction on the brain image and Stroke Scale scores, muscle strength scale scores, or initial diagnosis subtype.
Relation of Scan Findings to Assessments at 24 Hours
Patients were classified as to whether or not symptoms had resolved before 24 hours after stroke onset by use of NIH Stroke Scale scores at 24 hours. Figs 3⇓ and 4⇓ show the number of positive and negative CT and MRI scans at baseline and at 24 hours for patients with 24-hour Stroke Scale scores ≤1 versus scores >1.
Among patients whose symptoms had returned to normal within 24 hours, the baseline CT was negative for 21 of 26 (81%); for these same patients, MRI at baseline was negative for 12 of 22 (55%). Among patients who worsened, whose course remained stable, or who improved but did not return to baseline by 24 hours (Stroke Scale >1), the baseline CT was positive for 28 of 50 (56%); for these same patients, MRI at baseline was positive for 28 of 48 (58%). Overall, CT scan findings at baseline showed significant differences by Stroke Scale outcome (χ2=9.4l, P<.01), while MRI scan findings at baseline did not show significant differences (χ2=1.01, P=NS).
However, when CT and MRI scans were performed at 24 hours, the findings for both scan types were related in a manner similar to Stroke Scale outcomes (Fig 4⇑). In patients with Stroke Scale scores ≤1, CT scans done after 24 hours were negative for 14 of 21 (67%); MRI scans were negative for 10 of 21 such patients (48%). Among those with Stroke Scale scores >1 at 24-hour assessment, CT was positive in 30 of 40 (75%) and MR was positive in 33 of 38 (87%). Differences in frequency of positive and negative 24-hour scan findings by Stroke Scale outcome were significant for both CT (χ2=10.01, P<.01) and MRI (χ2=8.47, P<.01).
Compared with the baseline findings, both MRI and CT more conspicuously defined the lesion limits after 24 hours. Of 18 patients with CT lesion volumes computed for both baseline and 24-hour scans, lesion size increased for 10, decreased for 2, and remained unchanged (<±10% change) for 6. Of 15 patients with MRI lesion volumes for comparison, 9 increased, 3 decreased, and 3 remained unchanged. Changes were not accounted for by edema, which was present in both increased and decreased lesions. Of 3 hemorrhagic strokes imaged repeatedly on CT, 2 remained unchanged and 1 decreased in size.
The study presented here indicates that with the use of the technology available through 1991, we did not find a notable difference between CT and MRI in the earliest detection of ischemic lesions. Only one case showed a lesion within 1 hour, but many began to demonstrate the lesion within 2 to 3 hours, and by 24 hours the topography of the lesion was equally evident on both techniques. Increases or decreases in sizes of lesions occurred in only a few instances, arguing that few lesions altered their topography after their outlines were first evident.
This cohort represents the largest reported number of patients soon after stroke who were scanned this early with rapidly performed CT and MRI and who had several follow-up clinical examinations.
Some studies have reported on only one technique; in the case of CT, imaging done within 4 to 6 hours of onset of stroke showed early changes of infarction in one study in 25 of 36 patients.5 Drawing on some of the cases also included in the present study, Yuh et al11 found changes of brain MR signal in the first few hours, although by 24 hours almost 90% of the patients developing an infarction observable on MRI showed a signal change on the long TR sequences, compared to 50% sensitivity to signal changes on the T1-weighted images.
Some attempts have been made to compare the findings on the two techniques in the early stages of infarction. With the use of 0.15-T resistive magnets with T1- and T2-weighted techniques, early experiences with infarction, studied over varying periods of time up to 36 hours, did not show a major difference between CT and MRI in 16 patients.14 Infarction was seen earlier on MRI than CT in eight patients, with times of 2, 6, 7, and 10 hours after onset. The investigators were unable to differentiate infarction from perifocal edema by CT or MRI. In two patients the lesions were seen on T2-weighted images earlier than on T1-weighted images. Both CT and MRI showed large cortical infarcts better than smaller lesions of the basal ganglia and brain stem. The size of abnormal findings gradually developed, reaching a maximum by day 5 to 7, after which they gradually subsided and reached a stable size after approximately 2 months. In more modern studies, Bryan et al15 found that by the end of the first 24 hours the long TR sequences of MRI are significantly more sensitive to the detection of infarction than CT, which can reveal only 58% of the lesions. On follow-up studies these workers found that CT is positive in 82% of the patients, whereas MR is positive in 95% of the cases that will eventually develop a detectable infarction corresponding to the site of the clinical stroke. The follow-up studies revealed new lesions that were not detected on the first examination, and lesions already seen appeared to be larger and more conspicuous on the follow-up studies.
The differences of technique between CT and MRI could make comparison of these two techniques difficult, especially for lesions of the brain stem. Well-established smaller lesions, particularly in the basal ganglia and posterior fossa, may be seen by MRI and not by CT.16 Regions of abnormal signal intensity may involve areas of normal function as inferred by neurological examination.17 Contrast agents may shorten the time to image the early lesion on MRI: Imakita et al18 found MRI more reliably positive than CT, with lesions shown more clearly on MRI with contrast enhancement.
Beyond 24 hours, 30% of patients with an infarction visualized in the first 24 hours show an increase in size on the follow-up scan.17 For patients with transient ischemic attack, 86 showed lesions on MRI versus 42% on CT.19 In some studies MRI seemed to image the infarct lesion earlier than did CT scanning,20 and the lesion topography appeared larger, possibly because of edema.3 However, the issue of reversibility of the clinical deficit has not yet been clarified.
The newer techniques of spectroscopy21 and diffusion-weighted imaging,22 introduced long after this project had accumulated most of its data, have been described in rat models showing changes as early as 30 minutes after experimentally induced focal ischemia in the middle cerebral artery,23 a finding subject to experimental manipulation with neuroprotective agents.24 Thus far the applicability of MR spectroscopy techniques in humans has been demonstrated in only a small series of patients,25 and the literature on the utility of diffusion-weighted MRI is only just emerging.
These newer techniques offer the hope of identifying ischemic tissue at risk for infarction. The feasibility and utility of such studies have not yet been proven in a cohort of patients imaged in the hyperacute phase of stroke comparable to the studies reported here, but it can be anticipated that such studies will be carried out. The difficulties experienced in achieving the present study suggest that the more complex and demanding the image technology, the harder it will be to make it practical and applicable as a diagnostic tool in the setting of hyperacute stroke and clinical trials.
The results of our study, generated by now widely available technology, suggest that in the hyperacute setting of the first few hours after stroke, the extra efforts to define the area of ischemia by obtaining an early emergency conventional MRI scan rather than using the more readily achieved CT may not yield extra dividends.
This study was supported by contract NS 5-1001 of the National Institute of Neurological Diseases and Stroke. The contributions of the authors were as follows: The clinical neurological investigators were J.P. Mohr, T.K. Tatemichi, H.P. Adams, Jr, and J. Biller. The radiologist staff included S.K. Hilal, W.T.C. Yuh, E. Tali, H. Nguyen, and I. Mun. The support staff at the University of Iowa included K. Grimsman. The statistician at KAI was S. Hedges.
- Received October 19, 1994.
- Revision received January 25, 1995.
- Accepted January 25, 1995.
- Copyright © 1995 by American Heart Association
Wall SD, Brant-Zawadski M, Jeffrey RB, Barnes B. High frequency CT findings within 24 hours after cerebral infarction. Am J Radiol. 1982;138:307-311.
Tatemichi TK, Mohr JP, Rubinstein LV, Kase CS, Nichols FT, Price TR, Wolf PA. CT findings and clinical course in acute stroke: the NINCDS pilot Stroke Data Bank. Stroke. 1985;16:138. Abstract.
Bozzao L, Bastianello S, Fantozzi LM, Angeloni U, Argentino C, Fieschi C. Correlation of angiographic and sequential CT findings in patients with evolving cerebral infarction. AJNR Am J Neuroradiol. 1989;10:1215-1222.
Brant-Zawadski M, Pereira B, Weinstein P, Moore S, Kucharczyk W, Berry I, McNamara M, Derugin N. MR imaging of acute experimental ischemia in cats. AJNR Am J Neuroradiol. 1986;7:7-11.
DeLaPaz RL, Shibata D, Steinberg GK, Zarnegar R, George C. Acute cerebral ischemia in rabbits: correlation between MR and histopathology. AJNR Am J Neuroradiol. 1991;12:89-95.
Yuh WT, Crain MR, Loes DJ, Greene GM, Ryals TJ, Sato Y. MR imaging of cerebral ischemia: findings in the first 24 hours. AJNR Am J Neuroradiol. 1991;12:621-629.
Fiorelli M, Sacchetti ML, Toni D, Argentino C, Gori C, DiBasi C, Gualdi G, Fieschi C. Feasibility and usefulness of conventional spin-echo MR imaging in hyperacute ischemic stroke: a comparison with CT scan. Circ Metab Cerv. 1993;10:5-10.
Bryan RN, Levy LM, Whitlow WD, Killian JM, Preziosi TJ, Rosario JA. Diagnosis of acute cerebral infarction: comparison of CT and MR imaging. AJNR Am J Neuroradiol. 1991;12:611-620.
Salgado ED, Weinstein M, Furlan AJ. Proton magnetic resonance imaging in ischemic cerebrovascular disease. Ann Neurol. 1986; 20:502-507.
Monsein LH, Mathews VP, Barker PB, Pardo CA, Blackband SJ, Whitlow WD, Wong DF, Bryan RN. Irreversible regional cerebral ischemia: serial MR imaging and proton MR spectroscopy in a nonhuman primate model. Am J Neuroradiol. 1993;14:963-970.
Minematsu K, Fisher M, Li L, Davis MA, Knapp AG, Cotter RE, McBurney RN, Sotak CH. Effects of a novel NMDA antagonist on experimental stroke rapidly and quantitatively assessed by diffusion-weighted MRI. Neurology. 1993;43:397-403.