Normal diffusion-weighted MRI during stroke-like deficits

Methods.

We reviewed clinical and radiologic records of a consecutive series of patients referred for MRI study with the initial diagnosis of ischemic stroke between December 1994 and September 1997. For the current analysis, patients were chosen in whom 1) DWI was normal in the clinically suspected brain region and 2) a neurologist made a clinical diagnosis of acute stroke and documented the presence of an acute, focal neurologic deficit both before and after DWI was performed. Medical history and risk factor profile for ischemic stroke were carefully reviewed in each patient. An acute stroke-like neurologic deficit was described as a new and sudden onset of focal neurologic symptoms referable to an intracranial vascular territory and including any of the following, either alone or in combination: unilateral hemiparesis or unilateral hemisensory loss involving face, arm, or leg in any combination; hemianopia; cognitive dysfunction, including aphasia, alexia, memory loss, and neglect; any eye movement or cranial nerve dysfunction localizable to the brainstem; unilateral ataxia, dysmetria, or dysdiadochokinesia.

All patients underwent emergency evaluation by one of the five stroke neurologists (W.J.K., H.A., F.S.B., L.H.S., G.R.). EKG, blood cell count, electrolytes, glucose, and liver and renal function tests were performed in each patient; all had an initial brain CT to exclude intracranial hemorrhage. Neurologic examination was repeated by the same physicians after the MRI study to determine whether the deficit was still present. In patients with normal MRI, further evaluation with EEG, lumbar puncture, and toxicologic screening was performed to investigate other possible causes of acute focal neurologic deficit. All patients underwent at least one vascular study, including carotid Doppler ultrasonography, transcranial Doppler sonography, magnetic resonance angiography (MRA), or selective cerebral angiography. Echocardiography (ECHO) was performed in the majority of patients. Follow-up CT or MRI was obtained in most patients. All MRI studies were evaluated by the same staff neuroradiologist (P.W.S., G.A.S., R.G.G.).

All patients with focal neurologic deficit and normal DWI were stratified according to their final discharge diagnoses. Ischemic stroke was defined as an acute, persistent, focal neurologic deficit consistent with involvement of an intracerebral vascular territory, with a clinically appropriate ischemic lesion in the follow-up MRI or CT. Patients with persistent clinical deficits, but with a normal or absent follow-up imaging study, were classified as having cerebral ischemia when the symptoms strongly indicated a small vessel stroke and other possible diagnoses were excluded. Patients whose symptoms resolved completely within 24 hours, who had multiple risk or etiologic factors for cerebral ischemia, and who had no other more appropriate diagnosis were classified as having TIAs and were included in the cerebral ischemia group. Seizure was diagnosed in patients with a stroke-like neurologic deficit and one or more of the following: eye-witnessed seizures, epileptiform activity on EEG, recurrent similar spells, reduction in attack frequency by antiepileptic treatment. Migraine was considered the best diagnosis if one or more of the following were present: history of migraine, similar previous episodes, scintillating visual illusions, headache, prevention by antimigraine therapy.

MRI studies were performed on a 1.5-tesla GE Signa MR instrument (Waukesha, WI) with echo planar capabilities provided by Advanced NMR Systems (Wilmington, MA). Acquisition parameters for DWI included the following: repetition time (TR) 6,000 msec, echo time (TE) 118 to 154 msec, matrix 256 × 128, field of view (FOV) 40 × 20 cm, slice thickness 6 mm, 1-mm gap, 17 or 18 axial slices, maximum b value of 1,221 sec/mm², and diffusion gradients applied in three or six orthogonal directions at an effective strength of 15 mT/m. Trace DW MR images were generated by using methodology previously described.11 Other routine MRI sequences included the following: sagittal T1-weighted images (TR/TE = 650/16, matrix 256 × 192, FOV = 24 × 24 cm, slice thickness 5 mm, 2-mm gap, 1 number of excitations [NEX]); fast spin-echo proton images (TR/TE = 2,500/13 [effective], echo train length = 4, matrix 256 × 192, FOV = 20 × 20 cm, slice thickness = 5 mm, 1-mm gap, 1 NEX); fast spin-echo T2-weighted images (TR/TE = 4,200/96 [effective], echo train length = 8, matrix 256 × 192, FOV = 20 × 20 cm, slice thickness = 5 mm, 1-mm gap, 1 NEX). Hemodynamically weighted MRI (HWI) was performed according to methodology previously described.11 In this technique, a bolus injection of the contrast medium induces a transient signal drop in the brain, reflecting an increase in the rate of T2 relaxation proportional to the concentration of the contrast medium.12 From these concentration-versus-time data, maps of relative cerebral blood volume (rCBV), relative cerebral blood flow (rCBF), and relative mean transit time (rMTT) were generated using published analytic methods.13 Head CT was performed on GE Advantage instruments (Waukesha, WI). Parameters included 5-mm contiguous slices obtained by using 140 kV and 340 milliamp seconds.

All patients were followed carefully during the first 24 hours to determine the clinical course and then by daily neurologic examinations until discharge to estimate the duration of symptoms. MRI latency, i.e., the time between symptom onset and the MRI study, was determined in each case. All calculations were expressed as mean ± standard deviation.

Results.

Via our stroke log we estimated that in the period of this study, 755 patients presenting with acute stroke-like deficits had an abnormal DWI. Twenty-five patients had an initial diagnosis of stroke and sudden onset of focal neurologic deficits that persisted after a completely normal DWI study. Two others, with DWI abnormality in one brain region, were included in this study because they had normal signal on DWI in the brain region clearly responsible for ischemic clinical symptoms. There were 12 women and 15 men. The mean age was 71.9 ± 11.5 years, ranging from 45 to 93 years. Table 1 summarizes stroke risk factor profiles and findings on neurologic examination. Three patients had one, and 21 patients had two or more, of the major stroke risk factors: hypertension, diabetes mellitus, hyperlipidemia, smoking, previous embolic event, atherosclerosis, and atrial fibrillation. No stroke risk factor could be identified in three patients, although in one, ECHO showed ruptured cordae tendinae (Patient 4). Sixteen patients (59%) had either cardiac or arterial sources of cerebral emboli based on EKG, ECHO, carotid ultrasonography, or cerebral angiography.

MRI studies including DWI were performed between 0.5 and 38 hours after onset of stroke-like symptoms (mean 9.6 ± 9.2 hours) (table 2). Follow-up imaging studies were performed in 19 patients—CT in 4 and MRI in 15. Follow-up studies were done between 1 and 7 days after the first study, except for two patients who were scanned 2 and 7 months later.

Ischemic strokes.

Based upon the clinical course and follow-up imaging studies, the final diagnosis was cerebral ischemia in 17 of 27 patients (63%) who presented with stroke-like deficits without matching DWI abnormality (see table 2).

Patients with eventual cortical infarction.

In 3 of these 17 patients (nos. 1 to 3), follow-up CT or MRI showed a new infarct in cortical regions implicated by both clinical localization of the initial symptoms and increased rMTT on HI. Each had neurologic signs and symptoms indicative of involvement of brain regions that were normal on DWI but ischemic by HWI (figure 1). The HWI pattern in each of these regions was identical—normal rCBV but increased rMTT and decreased rCBF. Two patients (nos. 1 and 2) had additional regions with abnormal DWI and matched reduced rCBV on the initial scan. The third patient (no. 3) had symptoms consistent with the left middle cerebral artery ischemia, decreased flow signal in the M-2 branch of the middle cerebral artery on MRA, and regions of increased rMTT with decreased rCBF in the frontal and parietal lobes (see figure 1). Initial DWI was normal. Her deficits partially improved but later worsened again. A week later, follow-up CT study showed multiple, small, patchy regions of infarct in the left frontal and left parietal cortical regions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. Patient 1 presented with acute alexia without agraphia, color anomia, and right homonymous hemianopia. Initial diffusion-weighted MRI (DWI) showed an increased signal in the left hippocampus. Relative mean transit time (MTT) and relative cerebral blood flow (rCBF) maps suggested that the occipital cortex was ischemic. On the follow-up DWI 5 days later the occipital cortex and splenium were involved in the infarction. Patient 2 had left angular gyrus (partial Gerstmann’s) syndrome at admission not explained by the increased signal in the frontal cortex on DWI. Initial relative MTT and rCBF images showed decreased perfusion in the clinically symptomatic region that later was involved in the infarction seen on the follow-up DWI 5 days later. Patient 3 had fluctuating conduction aphasia, right hemiparesis, and right extinction to tactile stimulation. Initial DWI was normal, but relative MTT showed abnormal perfusion in the clinically relevant regions. Follow-up CT 7 days later revealed a small hypodensity in the left supramarginal gyrus.

Patients with lacunar syndrome or lacunar infarction.

Three patients with a classical, persistent, lacunar-type deficit and a negative initial MRI had lacunar infarction on the follow-up study (Patients 4, 5, and 6). Retrospective reassessment of the initial DWI in two of these (Patients 4 and 5) demonstrated a subtle region of DWI signal hyperintensity in the location of the lesion seen on follow-up imaging (figure 2). Three other patients had clinical symptoms attributable to a territory of small penetrator vessels with normal follow-up imaging (Patients 8, 9, and 10). In another, symptoms resolved over 3 days, and no follow-up study was obtained (Patient 7).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. Patients 4, 5, and 6 presented with a lacunar syndrome localizable to the brainstem. All had normal diffusion-weighted MRI initially. Small brainstem lacunes were seen on follow-up DWI studies performed 1 day later in Patient 4, 3 days later in Patient 5, and 2 days later in Patient 6. In retrospect, a small hazy area of increased intensity could be seen on the initial scans in Patients 4 and 5 in the clinically relevant regions.

Patients with reversible ischemic deficits and normal follow-up MRI.

Five patients (nos. 13 to 17) had a discharge diagnosis of TIA, with mean duration of symptoms of 13.4 ± 6.5 hours. Neurologic deficits lasted longer than 12 hours in four of five, and 4 hours in the fifth. Two other patients (nos. 11 and 12) had reversible cortical deficits that resolved after 24 hours of onset. In all patients with TIAs as well as those with longer-lasting but eventually resolving cortical deficits, there was partial improvement in neurologic symptoms by the time of the DWI study. None had HWI.

Discussion.

DWI promises to improve diagnostic accuracy over that achievable by clinical examination, CT, or T2-weighted MRI alone.9-11 In animal studies, DWI abnormality coincided with cellular energy failure.7,8,14,15 The major energy-dependent system in ischemic tissue is Na-K adenosine triphosphatase. Its activity decreases within 2 minutes of adenosine triphosphate decline.16 This causes intracellular sodium accumulation and resultant influx of water from the extracellular space. The extracellular space narrows by as much as 50%, thereby decreasing diffusibility of intercellular water.17 Furthermore, with a greater proportion of water intracellularly, the Brownian movement of tissue water is more restricted by the intact cell membrane18 and the gelatinous intracellular matrix. The subsequent decrease in the ADC of water appears as a bright signal on the DW MR images. The excellent concordance between DWI brightness and the tissue energy level makes this technique a reliable tool for detecting ischemic injury. However, the dynamic and highly variable nature of cerebral ischemia in humans raises rare but clinically important obstacles in accurate interpretation of DWI.

Our data describe a group of patients in whom DWI was normal in the clinically affected brain region during a stroke-like neurologic deficit. Despite extensive testing, accurate diagnosis in this group is difficult but includes stroke mimics (37% of our patients) such as seizure, migraine, primary CNS lymphoma, transient global amnesia, and functional disorder. Cerebral ischemia/infarction likely accounts for the etiology in the majority of patients (63% in our cohort). However, only 17% of them developed a major cortical infarction in our study. To our knowledge, this is the largest published series of patients with normal DWI at the time of clinical deficit.

Warach et al.19 reported 19 stroke patients studied with DWI and HI. They found DWI/HI to be very sensitive in predicting clinical outcome. However, 6 of their patients had normal DWI/HI, although 5 had been studied with only a single brain slice technique. This study underscores the importance of studying the pathophysiologic diagnoses in patients with stroke-like deficits and normal DWI using the multi-slice technique.

It is important to know that DWI can be normal in the ischemic brain regions that are causing clinical deficit. Our three patients (nos. 1, 2, and 3) with unremitting ischemic symptoms lasting at least 3 hours had normal DWI in the corresponding brain regions. In each, a pattern of increased rMTT and decreased rCBF but normal rCBV was seen on HI in the clinically affected brain regions (see figure 1). This is a less severe pattern than that characterized by abnormal DWI and reduced rCBV.13,20 In studies of acute stroke patients, the former pattern was not infrequently seen encircling regions abnormal on DWI or rCBV maps, or both, i.e., “penumbra around a core.”20 The absence of a central region of injury in each of the three patients above suggests that they in fact had “penumbra without core.” One possible explanation for this pattern is that compensatory vasodilatation occurs in regions with mild reductions in rCBF leading to an increase in collateral blood flow and maintenance of normal rCBV. Yet this reduced level of rCBF is not sufficient to maintain normal electrical activity and causes ischemic neurologic deficits.21 Because the threshold for cerebral infarction is dynamic, dependent upon both the degree and duration of ischemia, these regions may eventually undergo infarction if the duration of the ischemia becomes excessive or the degree of ischemia worsens.22,23 Indeed, in all three patients above, a portion of the symptomatic ischemic regions eventually was involved in the infarct on follow-up studies (T2/DWI or CT abnormality). Similarly, in acute stroke patients imaged early, the focal DWI/rCBV lesion can sometimes be seen to enlarge into the territory defined by increased rMTT and decreased rCBF.20

DWI may not show an ischemic lesion in some symptomatic patients with reversible ischemic deficits. This category was the largest group in our cohort, accounting for 53% of all patients with the final diagnosis of cerebral ischemia. Although each of our patients had strong circumstantial clinical evidence suggesting that the deficit was due to brain ischemia (see table 1), it was not possible without a sensitive measure of blood flow to be sure that there had been an ischemic etiology for their symptoms. We completely relied on clinical criteria and circumstantial evidence for diagnosis of TIA in this study. Similar to the hypothesis discussed directly above, in these patients the degree of ischemia, though sufficient to cause neural dysfunction, may not have been severe enough, or lasted long enough, to cause major energy failure and consequent DWI changes.7 Conversely, DWI abnormality could have been present at the height of symptoms and reversed with reperfusion before the imaging study. Firlik et al.24 reported that patients with rapidly resolving deficits had normal xenon-enhanced CT perfusion scans. All of our patients with negative DWI, but diagnosed as having reversible ischemic deficits, actually had some partial improvement in their deficits by the time of MRI studies. Mintorovitch et al.9 showed that in rats DWI abnormality reverted completely to normal if the arterial occlusion was released within 33 minutes. Minematsu et al.25 reported similar findings indicating partial reversal of DWI abnormality in insults lasting less than 2 hours. An alternative, but speculative, hypothesis is that neurologic recovery after certain sublethal ischemia-reperfusion injuries may occur over hours to days, analogous to the Todd’s paralysis that follows seizure. After transient ischemia, prolonged transmission failure at cortical synapses may underly motor dysfunction,26 a potential mechanism for such a clinical phenomenon.

Early DWI may not show a definite lesion in patients presenting with a lacunar syndrome. This might be especially evident in brainstem locations as all of our patients with initially missed lacunar infarctions had lesions in the brainstem. In accordance with our findings, 1 of the 6 patients with initially normal DWI in the series of Warach et al.19 had a small infarction on follow-up imaging. More recently, in a study of the diagnostic value of DWI in 39 patients with subcortical ischemic syndromes, Singer et al.27 reported 2 patients with normal DWI. However, it was unclear whether these patients had symptoms at the time of imaging. In our patients, an infarction may have been missed because its size might have been below the resolution of MRI. Incorrect slice selection, leading to partial volume effects, might have also been responsible for missing small infarctions. In some patients, the signal-to-noise ratio was insufficient to permit prospective identification of circumscribed, faint, early DWI abnormalities that could only be identified after comparison with the follow-up images.

Seizures with postictal deficits, brain tumors, and toxic-metabolic causes are the most common nonvascular conditions mimicking stroke.28 However, diagnosis of these conditions is not always easy. Elderly patients presenting with aphasia or confusion frequently raise a diagnostic struggle between ischemic stroke and stroke mimics. Early means of accurately diagnosing ischemic stroke are particularly critical in treatment decision-making, especially regarding IV thrombolytic drugs. The present study was not intended to analyze stroke mimics. However, our demonstration of an MRI-defined level of ischemia causing neurologic deficits but preceding infarction indicates that a sensitive and specific measure of normal rCBF is a necessary complement to a negative DWI in accurately differentiating a stroke mimic from a true stroke.

Classical migraine, especially the acephalgic form in the elderly, may resemble stroke.29 HI, but not DWI, abnormalities have been reported during the visual aura in migraine patients.30 Spreading depression-like episodes were associated with aura in studies by PET and SPECT.31,32 Such episodes are highly energy consuming and, in animal models, are associated with a remarkable but transient decrease in ADC.33,34 Our patients with a final diagnosis of possible complicated migraine did not have DWI abnormalities. However, transient waves of spreading DWI abnormality would almost certainly have escaped detection in our short scanning time window.

Because we did not review the extensive diagnostic and clinical data on the 755 patients with abnormal scans during the study period, the present study does not address the specificity of DWI for detecting cerebral ischemia. However, our experience strongly supports the diagnostic value of DWI in evaluating patients who present with acute stroke-like deficits.10 Patients with small lacunar infarcts may escape detection by DWI. Patients with epileptiform or migrainous causes of stroke-like deficits did not have abnormal DWI signal. Importantly, this study demonstrates that in occasional stroke patients, the excellent concordance between depleted tissue energy stores and DWI brightness may not faithfully echo the relationship between the clinical symptoms and brain ischemia. DWI may still be normal in some patients with ischemic neurologic deficits who later progress to have infarction in regions initially abnormal on rMTT/rCBF maps. In patients presenting with an acute stroke-like syndrome and normal DWI, a sensitive indicator of normal cerebral blood flow is necessary to reliably exclude an ischemic cause. In addition, low cerebral blood flow may also occur in some patients with migrainous aura30,31 or postictal cortical depression.35 Therefore, how to best use DWI in concert with MRI-based rMTT/rCBF maps as a clinically valuable indicator of significant brain ischemia requires further systematic study.