To investigate the development of ischemic brain lesions, as present in the acute stroke phase, by diffusion-weighted magnetic resonance imaging (DWI), and in the subacute and chronic phases until up to four months after stroke, in fluid-attenuated inversion recovery (FLAIR)- and T2-weighted (T2W) magnetic resonance (MR) images.
Materials and Methods
Twelve consecutive patients with their first middle cerebral artery (MCA) infarction were included. Lesion volumes were assessed on T2W images recorded with a turbo spin echo (TSE) and on images recorded with the FLAIR sequence on average on day 8 and after about four months. They were compared with acute lesion volumes in perfusion and DWI images taken within 24 hours of stroke onset.
On day 8, lesion volumes in images obtained with FLAIR exceeded the acute infarct volumes in DWI. The chronic lesion volumes were almost identical in T2W and FLAIR images but significantly reduced compared with the acute DWI lesions. The lesion volumes assessed on DWI images correlated highly with the lesions in the images obtained with TSE or FLAIR, as did the lesions in the images obtained with FLAIR and TSE. The secondary lesion shrinkage was accompanied by ventricular enlargement and perilesional sulcal widening, as most clearly visible in the images obtained with FLAIR.
HUMAN BRAIN INFARCTION is a dynamic process eventually leading to a circumscribed brain lesion. The initial interruption of the cerebral blood supply causing a complex pattern of perfusion, electrolytic, metabolic, and neurotransmitter changes is followed by secondary events such as edema formation, cytokine and gene expression, and blood cell infiltration (1). In the acute stage of stroke there are two primary clinical goals: First, it is essential to identify the volume of the perfusion deficit and of the ischemic brain lesion as early as possible to initiate medical treatment, including thrombolysis (2). Second, it is important to establish a prognosis concerning the imminent clinical deficit and the possibility for poststroke recovery. Since it is generally held that the extent of the brain infarct determines the clinical prognosis (3–5), detailed knowledge about the infarct lesion is fundamental for acute stroke management.
Magnetic resonance imaging (MRI) readily allows determination of ischemic brain lesions and thus has become an integral part of diagnostic procedures in the clinical evaluation of acute stroke. Most widely used in this setting is diffusion-weighted MRI (DWI), which has a high sensitivity to detect acute ischemic lesions (6, 7). DWI captures random movements of water protons, which are restricted in ischemic brain tissue due to cytotoxic edema and to an increase in cell volume (7–9). However, DWI does not show the area of impaired cerebral perfusion. This is directly accessible by perfusion imaging (PI). In fact, it was shown that in the acute stage of ischemic stroke the perfusion deficit area exceeds the DWI lesion, whereby the PI-DWI mismatch demonstrates the brain tissue potentially salvable by thrombolysis (2, 10, 11). Further, the acute DWI lesions are inhomogeneous and differ from normal tissue in various ways due to different temporal rates of lesion evolution toward infarction (12). In lesion follow-up, it has been shown that changes of diffusion due to acute ischemia become manifest between 24 and 48 hours after symptom onset and normalize after about 10 days (13, 14). Therefore, other MRI techniques are required for studying the long-term consequences of ischemia in the human brain beyond the acute phase. Usually, T2-weighted (T2W) images recorded with a spin echo sequence (TSE) have been used to assess the more long-standing stroke lesions. Recently, the fluid-attenuated inversion recovery (FLAIR) sequence producing T2 contrast-dominated images with suppression of the cerebrospinal fluid (CSF) signal has been shown to be more sensitive in the detection of acute brain lesions, including stroke (15, 16).
For the acute stage after stroke, a lot of information is available about lesion development of up to approximately eight days after infarction. For example, it has been described that human stroke lesions grow initially after the insult, as demonstrated with cranial computer tomography (CT) and MRI (17–19). Also, it was found that the acute PI and DWI changes predict lesion evolution (9, 18–22). It is, however, unclear in which way and to what degree a brain infarct develops beyond the acute stage. In particular, we wondered if a stroke lesion may regress in volume—and if so, whether there are accompanying changes in adjacent brain structures. The aim of this prospective study was therefore to investigate the chronic lesion development with a long follow-up of about four months for two reasons. First, we sought to assess whether the ischemic brain lesions in the subacute and chronic phases correspond to the DWI lesions 24 hours after stroke onset. And second, we wished to investigate if the infarct lesions are depicted similarly by the clinically most widely used FLAIR and TSE sequences.
MATERIALS AND METHODS
We examined the MRI scans of 19 consecutive patients (57 ± 14 years, range = 31–87 years) with their first acute completed ischemic stroke. According to our diagnostic protocol, which was approved by the Ethics Committee of our university, the patients were subjected to PI after intravenous gadolinium (Gd) bolus injection, DWI, and time-of-flight MR angiography, as detailed previously (22). Inclusion criteria for this study were that the patients received three sets of MR images: 1) a DWI image within 24 hours of symptom onset that showed an acute supratentorial territorial lesion in the middle cerebral artery (MCA) territory, 2) MR images recorded with the FLAIR and TSE sequence in the subacute stage about one week (6–13 days, average = 8 ± 2 days) after onset of cerebral ischemia, and 3) MR images recorded by the FLAIR and TSE sequence in the chronic stage after about four months (73–184 days, average = 124 ± 33 days). Furthermore, patients were excluded who had symptomatic or space-occupying secondary hemorrhage on the follow-up MRI scans. Twelve patients fulfilled these criteria (Table 1). The neurological deficit was assessed in parallel using the European Stroke Scale (23). Seven patients were subjected to systemic thrombolysis with recombinant tissue plasmingen activator (24), while five patients were not, due to spontaneous regression of the neurological deficit or delayed arrival in hospital.
Table 1. Characteristics of the Patients Investigated
MRI was done on a Siemens Magnetom Vision 1.5-T whole-body MR scanner (Siemens AG, Erlangen, Germany).
In the acute phase, PI and DWI images were recorded. The measurement of PI was performed by serial T2*-weighted single-shot echo planar imaging sequences, scanning 12 slices 40 times every two seconds. The parameters were TE = 54 msec, 12 slices with 5-mm thickness, interslice gap = 1.5 mm, matrix = 128 × 128, and field of view (FOV) = 240 mm. At the time of the fourth scan, 15 mL of Gd contrast agent (Magnevist®, Schering) was injected as a bolus at a rate of 5 mL/second, immediately followed by 15 mL of NaCl at the same rate. DWI was performed with a single-shot echo planar sequence with two b-values, b = 0 and 1000 seconds/mm2 in three dimensions in space: readout, phase encoding, and slice direction, resulting in four images per slice. The imaging parameters were TE = 100 msec, 20 axial slices, 5-mm thickness, interslice gap = 1.5 mm, matrix = 96 × 128, FOV = 240 mm, scan time = 26 seconds.
At two later dates, after about one week and about four months, the patients received T2W images using a turbo gradient spin echo sequence (TSE) with TR = 7040 msec, TE = 115 msec, echo train length = 69, 20 axial slices parallel to the base of the skull, thickness = 5 mm, interslice gap = 1.5 mm, flip angle = 160°, FOV = 230 mm, matrix = 345 × 512, and scan time = 1 minute 17 seconds. In addition, MR images were recorded with the turbo FLAIR sequence: TR = 9000 msec, TE = 105 msec, TI = 2200 msec, flip angle = 180°, 18 slices, thickness = 5 mm, interslice gap = 1.5 mm, matrix = 140 × 256, FOV = 240 mm, and scan time = 3 minutes 9 seconds.
Postprocessing and Image Analysis
All data resulting from PI and DWI were transferred to a SunUltraSparc1 (Sun Microsystems, Palo Alto, CA) workstation. From the PI data, time-to-peak parameter maps (TTP) were calculated. TTP here refers to the time expressed in seconds between the first scan and when the signal intensity reached its minimum produced by the bolus of contrast agent in the T2*-weighted series of images. By averaging the images obtained with the three different diffusion weighting directions (b = 1000 seconds/mm2), so-called diffusion-weighted trace images were calculated for each slice. The volumetric analysis of DWI trace images was performed as described previously and is briefly summarized here (25). A semiautomatic segmentation technique was applied using the softwares MPItool (Advanced Tomo Vision) and Matrox Inspector (Matrox Imaging). The average image intensity of healthy tissue in the nonaffected hemisphere was determined by two independent observers (A.R., S.M.) on regions of interest (ROIs) laid in the unaffected contralateral hemisphere. The average intensity of healthy tissue for each separate slice was then calculated by averaging intensities within the ROIs and thereafter averaging the measurements of the two observers. Thus, the variance of the mean values and the error propagation related to the establishment of the normal control image intensity were reduced. The interobserver error was 4%. To determine the volume of the acute infarction, a segmentation at a threshold of four seconds and 20% higher than the average intensity of the healthy tissue was applied automatically to the entire PI and DWI images, respectively. It was shown previously that a TTP delay of four seconds determines the neurological deficit and the ischemic brain tissue at risk, while the 20% threshold in the DWI images predicts infarct lesion volume at eight days (22, 25).
Since the MR images recorded with both FLAIR and TSE showed a wide range of image values for normal gray and white matter, no automatic image analysis procedure was applicable for lesion morphometry in these images. Therefore, volumetric analysis was carried out by manually tracing the outer edge of the hyperintense lesional tissue in the MR images by two independent observers using the software MPItool (Advanced Tomo Vision), as outlined previously (25). Criterion was the border at which the gray values of normal brain tissue exhibited a stepwise increase to those of the hyperintense infarct lesion. The observers were blinded to subject numbers as well as to instants of MRI measurements. There were no artifacts in the MR images of either sequence; all images of each sequence were thresholded identically among patients and scanning occasions.
Statistical analyses were descriptive using mean and standard deviations for demographic and image data. Interobserver errors were calculated from mean square deviations. Correlations between different sets of volume data were analyzed using the Pearson's product moment correlation coefficient. The mutual concurrence of data sets was probed by an analysis of variance (ANOVA) with posthoc paired t-test. Change of neurological status was assessed with the Mann-Whitney test. Additionally, quantitative lesion volumes were tested by examining volume ratios.
As illustrated in Fig. 1, the ischemic brain lesions showed remarkable qualitative changes from the acute to the subacute after one week and to the chronic phase after approximately four months. DWI readily showed the infarct lesion. In most patients, six image slices were affected; however, even the smallest DWI lesions extended over two image slices. As illustrated in Fig. 1, the acute DWI lesions were homogenous hyperintense areas suggestive of a local tissue swelling. In the subacute phase, the lesions in the MR images obtained with FLAIR and TSE were comparable to the acute DWI lesions. However, there was perifocal edema leading to a compression of the adjacent lateral ventricle. In contrast, in the chronic phase, the lesions in the images obtained with TSE and FLAIR seemed to be smaller than the acute DWI lesions. Furthermore, the ischemic lesions had turned into inhomogenous lesions in the images obtained with FLAIR at the chronic stage, which was not as clearly visible in the T2W-weighted images (Fig. 1). Typically, the lesion centers turned hypointense, while the lesion borders remained hyperintense. In addition, the images obtained with TSE or FLAIR showed a postischemic enlargement of the lateral ventricles, particularly on the affected hemisphere in the chronic phase. Also, the sulci in the cortex close to the infarcted area became widened, suggesting brain atrophy (Fig. 1).
The results of the volumetric measurements are summarized in Table 1. The acute lesions were quite variable in volume. The PI lesions exceeded the DWI lesions in virtually all patients (P < 0.01). Additionally, the PI lesions were larger in the patients subjected to systemic thrombolysis than in those not eligible for thrombolysis (P < 0.02), showing a profound PI-DWI mismatch of some 100 mL in the patients subjected to thrombolysis. This was paralleled by a more severe neurological impairment in the patients treated with thrombolysis before treatment onset (P < 0.001). In the subacute stage, the lesion volumes that were traced at the outer edge of the hyperintensity had increased significantly, compared with the acute DWI lesions in the images recorded with FLAIR. In contrast, they were on average slightly decreased in the T2W images recorded with TSE (Fig. 2). The chronic lesion volumes were reduced in both types of MR images (Table 1). It became evident that the chronic stroke lesions in the images recorded with FLAIR and TSE were significantly smaller than the acute DWI lesions (Fig. 2). Remarkably, all patients were neurologically improved (P < 0.01) at four months after stroke (Table 1). Correspondingly, in the patients treated with systemic thrombolysis, the lesions in the DWI, T2W, or FLAIR images were statistically not different from those not treated, although they were somewhat larger in the patients subjected to thrombolysis. Figure 3 shows that the ratio of the ischemic lesions in the patients treated with thrombolysis, compared with the nontreated patients, was greatest for the PI volumes and smallest for the chronic FLAIR lesions, but in general about twice as large for the DWI, T2W, and subacute FLAIR lesions.
When analyzing the lesion evolution longitudinally in each patient, it became apparent that there was a high correlation between the acute DWI lesions and the lesion volumes in the subacute and chronic stages, as evident in the images recorded with FLAIR (Table 2). Also, the DWI lesions correlated with the lesions in the T2W images obtained with TSE; however, this correlation was slightly lower (Table 2). Remarkably, the infarct lesions correlated in either type of MR images, in both the subacute and chronic stages (Table 2). In contrast, neither in the treated nor in the nontreated patients was there a correlation of the lesions in the PI images with the subacute or chronic MR lesions (Pearson correlation coefficient < 0.5). However, there was a weak correlation of the PI lesion with the acute neurological score (Pearson correlation coefficient = 0.58) and of the chronic lesions in the MR images recorded with FLAIR with the chronic neurological score (Pearson correlation coefficient = 0.64), in agreement with earlier observations (3–5, 22, 25). Testing of the interobserver error showed errors of 3 and 5 mL for the MR-weighted images recorded with FLAIR and TSE, respectively.
Table 2. Evolution of the Brain Infarct Volumes
Scanning time points
MR-sequence at each Instance
Correlation coefficient (r2)
Difference P (t-test)
DWI / FLAIR
DWI / T2W-TSE
FLAIR / FLAIR
T2W-TSE / T2W-TSE
DWI / FLAIR
DWI / T2W-TSE
FLAIR / T2W-TSE
FLAIR / T2W-TSE
In this study we investigated the volumetric development of the ischemic brain lesions with MRI until about four months after stroke onset. We were interested in whether the acute DWI lesions predicted the resulting chronic infarct and if the infarcts changed in volume over time. In this small sample of patients with their first ischemic stroke, we used DWI trace images acquired in the acute stage within 24 hours of stroke and MR images acquired in the subacute phase (about day 8) and the chronic phase (about four months after stroke) by the FLAIR and TSE sequences. Most remarkably, lesion morphometry showed diminished lesion volumes in the chronic phase in the images obtained with both sequences (Table 1, Fig. 2), suggesting lesion shrinkage. That this actually was the case is illustrated in Fig. 1: already the qualitative analysis of the MR images showed that the infarct was largest in the acute DWI images and subacute MR images obtained with the FLAIR and TSE sequences, regressing in volume thereafter. The primary focus of this study was not the PI-DWI mismatch, which is considered the target for thrombolysis (2, 10, 11). Nevertheless, this small sample of consecutive patients suggests that thrombolysis can salvage a large portion of ischemic brain tissue, which is reflected by a profound neurological improvement (Table 1). This pathophysiologically important issue is the subject of a recent systematic study on a larger group of patients (26).
Although it has been appreciated since the advent of neuroimaging that patients who have suffered from brain infarction exhibit residual brain changes, they have not for long been assessed in a systematic, quantitative manner. Only recently, it was described that human stroke lesions grow initially after the insult, as demonstrated with CT and MRI (17–19). This can be partially due to the fact that acute ischemic stroke may show some degree of hemorrhagic inhibition, which is clinically completely nonapparent but may result in transient mass expansion approximately one week after stroke, secondary to local vasogenic edema. Notably, patients with symptomatic or large secondary hemorrhage were excluded from this study. The subsequent lesion shrinkage shown in this volumetric study confirms and extends similar observations by others (13, 19, 20). These and our data accord in showing that postischemic lesion evolution extends to a long period of months beyond the acute stage of stroke. Specifically, the subacute lesions as depicted by the FLAIR sequence were larger than those in the acute phase as assessed by DWI. The increase of about 26% on average, as evident from the images obtained with FLAIR, is comparable to the volume increase within the first week of stroke as described by Lansberg et al (12), who studied the ADC, and Beaulieu et al (19), who employed FLAIR DWI. But it is significantly larger than the identical lesion, as determined in the T2W images obtained with TSE at the same time (Fig. 2). From a technical point of view, it can be said that the infarct lesions, including the perifocal edema, could be traced with greater certainty and interobserver reliability in the images obtained with FLAIR than in the images obtained with TSE. In fact, a confounder in volumetry of the T2W images obtained with TSE was that the MR signals in the stroke lesions and in the CSF were indistinguishable, while the CSF in the images recorded with FLAIR was black. Therefore, it was difficult in some cases to differentiate a cortical lesion at the outer brain surface from the CSF in the cerebral sulci. This probably resulted in a more conservative and smaller outline of the lesion in the T2W images obtained with TSE. Since the lesion volumes determined from T2W images in the subacute stage correlated with the lesion volumes in the acute DWI images and with those in the chronic T2W images (Table 2), the observed changes seem to indicate a systematic volume change.
The increase in lesion volumes between 24 hours after stroke and the subacute phase at about day 8 can be explained by development of secondary changes, including vasogenic edema and possibly leukocyte infiltration (1), which may contribute to the lesion volume as assessed with the FLAIR sequence. Thus, images obtained with FLAIR may be particularly suitable to assess the lesion volume in the subacute phase after stroke. In the chronic phase, the infarct lesions were significantly reduced in volume, partially due to lesion consolidation with dismantling of necrotic brain tissue and the resulting pseudocystic tissue change. It is well possible that this lesion shrinkage continues even beyond four months after stroke onset. The identical lesion volumes in the T2W images obtained in the chronic stage indicate that our lesion measurements were highly reliable (Fig. 2). Thus, both types of images turned out to be similarly useful for quantitative lesion assessment in the chronic phase after stroke. It should be cautioned, however, that the acute DWI image may not represent the true volume of the brain infarct, but instead may overestimate it. The reason may be suggested by Fig. 1, which shows some degree of lesion swelling resulting in compression of the lateral ventricle even at this acute. Thus, the dynamic processes initiated by brain infarction (1) may affect lesion volume measurements in MRI in all stages after ischemic stroke.
Figure 1 shows that in addition to the changes of the ischemic brain lesion itself, also the adjacent brain structures seem to shrink after stroke. It is well established in children that brain atrophy may develop after brain infarction, affecting the cerebellum and thalamus (27, 28). These studies support the neuroradiological observations that sulcal widening and ventricular enlargement may follow ischemia and that ischemic white matter damage also involves the corpus callosum (29–31). In this study, however, we did not determine the perilesional or remote structural changes after stroke such as thalamic atrophy or cortical thinning (32, 33). Such measurements are important for assessing the complete effect an infarct bears on the brain. In contrast to the stroke lesion itself (34), these measurements cannot be done adequately on conventional two-dimensional MR images but require volumetric three-dimensional MR images. Indeed, voxel-guided morphometry provides objective means to assess lesion shrinkage and the accompanying enlargement of the intra- and extracerebral fluid spaces (35). By this technique, the ispilesional thalamus, the homologue cortex in the contralesional hemisphere, and the contralesional cerebellum were shown to be atrophied after ischemic brain infarction (36). Consequently, there seems to be atrophy of brain tissue that is heavily connected with the infarcted part of the brain.
Our morphometric data are of relevance for clinical studies employing lesion volumes for assessing possible effects of therapeutic interventions. In accordance with recent studies (2, 10, 11), thrombolysis salvages brain tissue at the risk of resulting in smaller brain lesions than anticipated from the acute PI lesions (Table 2). Nevertheless, we have shown recently that DWI lesions beyond four hours after stroke onset most accurately determine the infarct volume on day 8 (25). However, a secondary growth of infarct lesions after thrombolysis was related to the degree of ischemia and the presence of diabetes mellitus (37–39). In addition, there are profound structural and functional changes beyond the infarct lesion (35, 36). Our data suggest that not simply the acute DWI lesion volume, but rather the lesion volume and location obtained with the sensitive FLAIR sequence predicts the degree of poststroke recovery and should favorably be used for morphometric-clinical correlation in stroke patients (5, 40).
The authors thank Erika Rädisch for expert technical assistance during MR scanning.