Significance of hyperintense vessels on FLAIR MRI in acute stroke

Patients and methods.

We retrospectively reviewed 358 MRI scans of 304 patients with stroke admitted between August 1997 and April 1998. Scans of patients with intracranial hemorrhage were excluded (n = 54). The remaining scans were from 198 men and 106 women (age range, 35 to 89 years). A stroke neurologist selected all MRI scans including all patients with HVS irrespective of their diagnosis. Two fellowship-trained neuroimagers blinded to the clinical information reviewed the MRI scans. Reviewers were aware of the significance of HVS but were not aware if HVS cases were included. MRI scans were reviewed randomly and individually. After the first review, both observers were given 90 MRI scans including 30 age-matched normal scans, 30 with parenchymal changes, and 30 HVS scans in a random order. The reviewers filled out a form noting the presence, location, and grading of HVS. The time to MRI was defined as the interval between the onset of symptoms and initial MRI examination (rounded off to the closest hour). HVS was identified as tubular hyperintense signal relative to gray matter on FLAIR, and on postcontrast T1WI, as vascular enhancement in two or more vessels. HVS was graded as follows.

Results.

A total of 304 MRI scans were reviewed, of which 15% (n = 46) were normal. Lacunar infarcts accounted for 19% (n = 58) of the strokes. HVS was identified on 10% (n = 30) of stroke MRI scans. The interobserver agreement for the identification and grading of HVS was substantial. Observer 2 identified all HVS patients, Observer 1 missed one, and Observer 2 graded a grade III as II. On the second review, both observers identified all HVS patients and graded them correctly. Twelve patients were grade I, 13 grade II, and 5 grade III.

All patients with HVS underwent MRI examination within 24 hours of onset of stroke symptoms (table). The age of the patients with HVS ranged from 52 to 81 years; 12 women and 18 men were included. In three patients who were rescanned 24 to 36 hours postictus, HVS was still present with large vessel occlusion and patients were symptomatic. HVS was seen in both complete and partial MCA distributions (figure 1). HVS was observed in both cortical and subcortical infarcts. In subcortical infarcts, HVS was located at the adjacent brain surface. Angiographic studies in patients with subcortical infarcts showed slow flow in leptomeningeal collaterals that ran from surface to the infarcted region. HVS was not seen with lacunar infarction possibly due to the small size of the affected vessels. With ICA or proximal MCA stenosis or occlusion, HVS was most intense and prominent in the sylvian fissure (figures 1 and 2⇓). Postcontrast T1WI obtained in nine patients with HVS showed intravascular enhancement in vessels that were hyperintense on FLAIR (figure 2, D and E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1. (A) Fluid-attenuated inversion recovery (FLAIR) image shows hyperintense vessels sign (HVS) in left sylvian fissure. (B) On SPECT, the area of hypoperfusion corresponds to HVS on FLAIR (A). (C) Three-dimensional time of flight MR angiography (3D-TOF), absent flow signal in left internal carotid artery and middle cerebral artery.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2. (A, B) Axial fluid-attenuated inversion recovery (FLAIR) images show hyperintense vessels sign (HVS) in the right middle cerebral artery (MCA) and anterior cerebral artery (ACA) distribution. (C) Three-dimensional time of flight MR angiogram (3D-TOF), absent flow signal in the right internal carotid artery (ICA), MCA, and ACA. (D, E) Postcontrast T1-weighted image shows intravascular enhancement in the vessels that are hyperintense on FLAIR. (F) SPECT; area of hypoperfusion corresponds to HVS on FLAIR. (G, H) Repeat FLAIR 2 days later shows infarction involving the right MCA and ACA territory; hyperintense vessels are still seen posterior to the MCA infarct (arrow).

Three patients studied within 4 hours of ictus demonstrated only HVS and no parenchymal, FLAIR, T1, or T2 changes on MRI. The FLAIR sequence was superior to TSE–T2WI in detecting ischemic involvement of the cortical ribbon. In five patients imaged within 6 hours of ictus, hyperintense signal on FLAIR involving the cortical ribbon or deep white matter in addition to HVS preceded changes on T2WI or T1WI.

All patients with HVS (n = 30) had large vessel occlusion or severe stenosis (>90%) either proximal or distal to the circle of Willis (figures 1 and 2⇑). There were 9 carotid occlusions (left 5, right 4), 9 MCA occlusions (left 3, right 6), 4 severe (>90%) carotid stenosis (left 2, right 2), and 17 MCA stenosis/poor flow (left 9, right 8). All carotid occlusions were verified by cerebral angiography (n = 8) and phase contrast MRA (n = 9). In patients who underwent angiography, HVS was always associated with angiographic evidence of slow flow. Of the stroke patients who had MRI in the first 24 hours, 45% (30/66) had HVS. Six patients without HVS also underwent cerebral angiography that confirmed large vessel disease (carotid occlusion 4, MCA occlusion 2) and demonstrated good leptomeningeal collaterals and absence of slow flow.

The average TCD velocity associated with HVS was 30% of the asymptomatic contralateral MCA. SPECT in 16 patients demonstrated moderate to severe hypoperfusion in the areas of HVS (figures 1B and 2F⇑). All patients with HVS were symptomatic at the time of MRI examination. The neurologic symptoms correlated with area of HVS. The patient in figure 2 presented with HVS on FLAIR in the right MCA and ACA distribution with a history of recurrent TIA. Forty-eight hours later, he was admitted with a large stroke involving ACA and anterior division branches of the right MCA. HVS was still present in the posterior division MCA branches and posterior ACA territory. The patient in figure 1 demonstrated increasing MCA velocities that correlated with better flow signal on MRA and improvement in NIHSS from 23 to 13. NIHSS ranged from 8 to 23 at the time of admission. In eight patients with clinical improvement, the NIHSS improved with increasing serial TCD velocity measurements.

On follow-up MRI (n = 14) and CT (n = 16) the infarct zone was clearly visualized. Twenty-six patients developed cerebral infarcts. The area of infarct was smaller than the area under hyperintense vessels in 21 patients (see figure 2). Five patients developed deep white matter infarction with ipsilateral >90% carotid stenosis or occlusion.

Discussion.

The FLAIR technique produces heavily T2-weighted CSF nulled images by coupling an inversion pulse followed by a long TI to a long TE. With nulling of CSF, areas of long T2 (hyperintense), especially in the cortex, gray–white matter interface, and periventricular region, become conspicuous. A number of studies have established the superiority of FLAIR over T2WI in lesion detection.12-14

We have described HVS on FLAIR sequence as an indicator of reduced flow and early ischemia and compared it with intravascular enhancement on postcontrast T1WI. Several reports have described intravascular enhancement as the earliest sign of ischemia.15,16 Our data suggest that HVS is the FLAIR counterpart of intravascular enhancement observed on contrast-enhanced MRI in patients with severe stenosis (90%) of large vessel or occlusion (figure 2). HVS was most prominent over the cerebral convexities and in the sylvian fissure. Larger arterial size and low signal from surrounding CSF is probably the reason. The flow speed depends on both pressure and the diameter of the lumen. Because of Bernoulli effect, with the same perfusion pressure a narrow vessel will have a higher flow speed than a larger one, thus explaining the greater likelihood of observing HVS in proximal vessels.

Our study suggests that HVS is the result of an arterial slow flow phenomenon. The angiographic observation of slow antegrade and leptomeningeal collateral arterial flow supports the idea that HVS is the result of slow flow. This finding may help differentiate arterial from venous strokes. The angiographic slow flow was substantiated by low TCD velocity measurements as well. Angiographic slow flow has been described with intravascular enhancement.5,11

To explain the origin of HVS it is important to review the factors responsible for the appearance of flowing blood on MRI. In normal cerebral circulation, fast moving blood, time of flight, and spin dephasing result in loss of signal in the blood vessels, resulting in flow void phenomenon. We presume that sluggish blood flow in the regionally dilated blood vessels near ischemic or infarcted tissue results in the loss of flow voids and vessels appear hyperintense against a dark CSF background.4,17-21 Alternatively, flow-related enhancement could add to the observed FLAIR signal.

HVS was observed in a small number of patients with stroke. Why HVS was not seen in all patients with stroke and whether HVS is related to the time of MR examination after stroke are not entirely clear. Some patients without HVS were scanned within the time period for HVS. In patients without HVS, we presume either collateral circulation compensated, large vessels had recanalized, or there was no slow flow, as noted in our non-HVS patients who underwent angiography. Angiography in these patients showed good leptomeningeal collaterals. The resolution of arterial enhancement on postcontrast T1WI is thought to represent the re-establishment of fast flow via recanalization or neovascularization.22 We believe the same mechanism is responsible for not observing HVS in all stroke patients.

Our observations suggest a correlation between the extent of HVS and the patient’s clinical condition. NIHSS scores were higher in patients demonstrating HVS involving the full MCA or more. In patients with resolution of HVS, improvement of TCD velocities accompanied improvement in NIHSS. The extent of HVS was more than the size of infarction on FLAIR images. It remains to be seen if the extent of HVS could be used as an independent predictor for assessing stroke size and severity.