Fluid-Attenuated Inversion Recovery (FLAIR) for Assessment of Cerebral Infarction
Initial Clinical Experience in 50 Patients
Background and Purpose Our aim was to evaluate fluid-attenuated inversion recovery (FLAIR) sequence in the diagnosis of cerebral infarction with MRI.
Methods A retrospective review was undertaken of 50 consecutive MRI studies ordered for suspected cerebrovascular accident. All studies included FLAIR and rapid acquisition with relaxation enhancement (RARE) T2-weighted spin-echo sequences. The two sequences were compared independently by four observers at two different institutions. Detectability of lesions and image quality were scored.
Results Overall, FLAIR sequences proved superior in 10 patients, detecting acute cortical infarcts missed with RARE spin-echo technique in five patients. In five additional patients, improved characterization of chronic infarction and improved detection of microangiopathic deep hemispheric changes were observed. One brain stem infarct was missed with the FLAIR sequence.
Conclusions FLAIR offers advantages in detection of acute infarcts affecting the cortical ribbon, is a useful, rapid adjunct to conventional T2-weighted spin-echo sequences, and has the potential to replace these in the future.
The detection of cerebral infarction influences patient management decisions. CT brain scanning can be negative in a substantial proportion of patients in the early stages of cerebral infarction.1 2 3 MR imaging of the brain has superior capabilities for the detection of cerebral infarction compared with CT4 5 6 7 but may miss hyperacute infarcts.8 9 When present, infarction is best depicted on MRI with T2-weighted SE sequences. With conventional and RARE sequences, cortical infarcts in particular can be hard to detect given the similarly high signal of cortical gray matter and adjacent CSF and the complex convolutional geometry of the surface of the brain.
FLAIR and its variations10 11 12 13 are MRI sequences that produce both a strongly T2-weighted image and suppressed CSF signal. To accomplish these twin goals, a conventional SE sequence is prefaced by a 180° inversion pulse. A relatively long TI is used to allow the longitudinal magnetization of CSF to return to the null point before SE imaging. Thus, the CSF signal is completely suppressed for cortical or periventricular areas, and lesions with typical T2 prolongation in the brain that are adjacent to spinal fluid become much more conspicuous compared with conventional T2 imaging.
The concept of long TI, CSF-nulled FLAIR relies on the known lengthy T1 magnetization recovery for CSF. With TRs of 10 or more seconds, such FLAIR images have been burdened by long scan times and poor anatomic detail. By coupling the concept of a 180° inversion pulse and long TI with a RARE SE readout module, the FLAIR imaging process can be accelerated into a time frame more amenable to routine clinical whole brain imaging.
The purpose of this report was to evaluate the accelerated FLAIR technique in patients with cerebral infarction. Two variations resulted from the implementation of the technique on two different MR systems at our site. The purpose of the study was not to compare implementations but rather to compare the FLAIR technique with more conventional T2-weighted images for the assessment of cerebral infarction.
Subjects and Methods
For the 6-month period before this study, a Turbo FLAIR imaging protocol had been added to our routine brain MR protocol. At our institution, this routine brain imaging protocol also contained at the minimum a T1-weighted sagittal sequence and a RARE SE axial sequence. We retrospectively selected 50 patients from our logbook who had brain MRI obtained during this period for suspected brain cerebrovascular accident. The FLAIR images were read independently from the turbo SE and fast SE acquisitions by four readers blinded to the exact clinical history (although they were told patients were scanned to rule out “stroke”) at two separate institutions.
Initially, the FLAIR acquisitions were obtained on a conventional Siemens 1.5-T VISION system (Iselin) with a standard circularly polarized head coil. The following parameters were used: TR=8000 ms, TI=2400 ms, ETL=7 with a 15-ms interecho delay, and nominal TE=105 ms. A 5-mm slice thickness was used. The matrix of 182×256 provided pixel measurements of 1×0.8 mm given the 183×210-mm field of view. Study time for this FLAIR was 3 minutes, 28 seconds. Toward the end of the study, our General Electric Signa 1.5-T system became capable of acquiring fast FLAIR images.13 The parameters on this instrument varied somewhat as follows: TR=11 000 ms, TI=2600 ms, ETL=7 with a 15-ms interecho delay, and nominal TE=146 ms. Slice thickness of 5 mm and matrix of 192×256 provided pixel measurements of 1×0.86 mm given the 220×220-mm field of view. Study time for this version of FLAIR was 6 minutes, 5 seconds. Our RARE sequences used the following parameters. For the Siemens VISION, the turbo SE protocol contained the following: TR=3500 ms, TE=22 and 90 ms, ETL=5, slice=5 mm, and matrix=192×256 over a 183×210-mm field of view. Flow compensation was applied along the read (anteroposterior) direction. For the General Electric Signa, the fast SE sequence was used, with TR=3500 ms, TE=22 and 90 ms, ETL=5, slice=5 mm, and matrix=192×256 over a 183×210-mm field of view. Flow compensation was applied along the read (anteroposterior) direction. The FLAIR and T2 protocols were acquired axially with phase encoding directed from right to left.
The readers evaluated the images in the following fashion. First, the particular scan being evaluated (FLAIR or RARE SE) was judged as normal or abnormal. If abnormal, the lesions were relegated to either the category of nonspecific patchy white matter lesions or focal, discrete infarcts including at least a small portion of the cortex. The readers were also asked to judge the conspicuity of any lesion(s) on a scale of 1 to 4 (equivocal, subtle, definite, “can't be missed”). The degree of artifact on the images was scored by the readers on a scale of 0 to 3 (none, mild, moderate, severe). Finally, overall diagnostic quality of the sequences was evaluated on a scale of 1 to 3 (poor, acceptable, excellent).
A total of 50 patients' scans were evaluated. Twelve patients had either completely normal studies or demonstrated only patchy white matter lesions on their studies. Thus, 38 patients had definable cerebrovascular lesions. Two patients demonstrated parenchymal hemorrhages (1 old, 1 new). The other 36 showed either old or acute infarcts. Overall, FLAIR sequences proved superior in 10 patients. In 5 patients, FLAIR sequence demonstrated infarcts missed with RARE SE technique. In 2 of these, small cortical gyral infarcts were seen with FLAIR but missed by three of the four readers with the RARE SE sequence (Fig 1⇓). In 2 others, territorial branch infarcts were seen only on the FLAIR images. One of these was a 6-hour left temporal infarct (posterior branches of the middle cerebral artery distribution) in a patient who demonstrated a second, inferior temporal infarct on both FLAIR and fast SE images (Fig 2⇓). Three of four readers missed the peripheral branch infarct on the fast SE images. The other was an infarct 4 hours old, in a patient with vertebral dissection and resulting posterior inferior cerebellar artery insult. Again, three of four readers missed this infarct on the fast SE images. In the fifth patient, FLAIR imaging demonstrated a watershed infarction to all four readers, whereas the RARE SE sequences suggested only patchy white matter lesions (Fig 3⇓). The additional visualization of the anterior and posterior cortical components of the infarction verified the presence of a watershed distribution infarct. Although in retrospect, particularly when presented next to the FLAIR images, the first fast SE image does show subtle changes in the affected cortex, the conspicuity of the lesion is much more apparent on the FLAIR image. In addition, in 5 other patients FLAIR imaging was better able to depict infarcts as old, whereas the RARE SE sequences demonstrated lesions indeterminate for age. This advantage of FLAIR sequences was based on the depiction of bulk water (spinal fluid signal) within the abnormal region (Fig 4⇓). Such characterization is, of course, aided by additional T1-weighted sequences (routinely done in the sagittal plane in our study).
Two infarcts were missed by FLAIR but shown on the RARE SE images. Both were in the brain stem. A right pontine lesion was missed on the FLAIR sequence (Fig 5⇓). A small (8-mm) left ventral medullary infarct was missed by two observers, in retrospect because of considerable CSF pulsation artifact present at the level of the foramen magnum (Fig 6⇓). An acute basal ganglial hemorrhage was readily identified as such on both FLAIR and RARE SE sequences (Fig 7⇓).
Conspicuity of lesions was judged better on the FLAIR sequences; an average score of 4 for FLAIR imaging versus 3 for the fast SE images was found. The degree of artifact was slightly greater on FLAIR images, but overall image quality was scored as essentially equivalent.
FLAIR imaging provides an advantage over more conventional T2-weighted imaging in that it nulls the signal from spinal fluid while providing a heavily T2-weighted image of brain parenchyma. The greatest benefit from this sequence in our patient cohort was the detection of cortical gray matter infarcts. The cortical ribbon in particular is quite sensitive to ischemia because of the high metabolic activity of this tissue, resulting in an early manifestation of its selective vulnerability. Because it is immediately adjacent to the spinal fluid present within the sulci, the cortex can easily be hidden as a result of partial volume artifact with the spinal fluid when imaged with sequences that emphasize the fluid signal. Not surprisingly, the greatest difficulty with such partial-volume artifact occurs when the cortical lesion is in the plane of the fluid space. Because cortical gray matter involvement is particularly isolated in the earliest stages of infarction, FLAIR offers advantages in the evaluation of acute infarction over RARE SE sequences. Although in most of our cases the “missed” infarcts could be detected by at least one reader on the first echo of the fast SE pair, it is clear that such infarcts, especially when acute, are much more readily apparent to all observers on the FLAIR images.
FLAIR also improved characterization of infarcts as chronic, particularly those that exhibit cystic encephalomalacia. Because the bulk water in such chronically infarcted regions has essentially the same makeup as CSF, its signal is nulled as well. It can thus be easily distinguished from solid tissue with increased water content (as seen in more acute infarction).
Given the relatively long TE of the FLAIR sequence, it is not surprising that a hemorrhagic lesion was readily detected (as it is with fast SE sequences) and distinguished from ischemic insult (Fig 7⇑). The long TE allows for considerable dephasing of spins due to the effects of deoxyhemoglobin in acute hemorrhagic foci and hemosiderin in old hemorrhagic foci. However, such a focus of low intensity from blood by-products could simulate a focus of CSF were it not for the surrounding edema and mass effect.
The failure of FLAIR to depict the midpontine infarct in our series was surprising, particularly in view of the touted advantages of FLAIR for white matter lesions.10 14 15 This case points to one potential weakness of this sequence; namely, it is a single-echo sequence: only one sampling of signal intensity is obtained. In a fortuitous instance, a lesion with a very long T1 (but not quite as long as CSF) and a long T2 can potentially find itself at the “crossover point” with normal tissue on a single-echo sampling. Thus, a dual-echo FLAIR sequence might offer a potential solution to this problem. Also, our initial implementation of the FLAIR sequence lacked gradient moment nulling. This accentuated spinal fluid pulsation artifacts, particularly at the base of the brain, leading to another “miss” when a medullary lesion was in the midst of ghost pulsated artifacts (Fig 6⇑). Improved sequence design to provide dual-echo capability and “flow compensation” has been undertaken.
In summary, our initial experience suggests that FLAIR is a useful adjunct to more conventional T2-weighted SE evaluation of the brain, particularly in the setting of cerebral infarction. The combination of T1-weighted sagittal, RARE axial, and FLAIR axial sequences takes less than 10 minutes. Indeed, with sequence improvements, the FLAIR sequence may replace conventional dual-echo T2-weighted SE imaging for routine brain screening with MRI.
Selected Abbreviations and Acronyms
|ETL||=||echo train length|
|FLAIR||=||fluid-attenuated inversion recovery|
|RARE||=||rapid acquisition with relaxation enhancement|
Reprint requests to Michael Brant-Zawadzki, MD, FACR, Hoag Memorial Hospital, 301 Newport Blvd, PO Box 6100, Newport Beach, CA 92658-6100.
- Received January 23, 1996.
- Revision received February 27, 1996.
- Accepted March 18, 1996.
- Copyright © 1996 by American Heart Association
Wall SD, Brant-Zawadzki M, Jeffrey RB, Barnes B. High frequency CT findings within 24 hours after cerebral infarction. Am J Radiol. 1982;138:307-311.
Brant-Zawadzki 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.
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.
Bryan RN, Levy LM, Whitlow WD, Killian JM, Preziosi TJ, Rosario JA. Diagnosis of acute cerebral infarction: comparison of CT and MRI imaging. AJNR Am J Neuroradiol. 1991;12:611-620.
Mohr JP, Biller J, Hilal SK, Yuh WT, Tatemichi TK, Hedges S, Tali E, Nguyen H, Mun I, Adams HP Jr. Magnetic resonance versus computed tomographic imaging in acute stroke. Stroke. 1995;26:807-812.
Alberts MJ, Faulstich ME, Gray L. Stroke with negative brain magnetic resonance imaging. Stroke. 1992;23:663-667.
De Coene B, Hajnal JV, Gatehouse P, Longmore DB, White SJ, Oatridge A, Pennock JM, Young IR, Bydder GM. MR of the brain using fluid-attenuated inversion recovery (FLAIR) pulse sequences. AJNR Am J Neuroradiol. 1992;13:1555-1564.
Hajnal JV, De Coene B, Cowan FM. Application of prolonged inversion and echo time inversion recovery (PIETIR) pulse sequences to diagnosis of disease of the brain. In: Book of Abstracts: Society of Magnetic Resonance in Medicine. Berkeley, Calif: Society of Magnetic Resonance in Medicine; 1992:1007. Abstract.
den Boer JA. Salverda P, Peters TR, Prevo RL. Multislice turbo-FLAIR in brain studies of multiple sclerosis. In: Book of Abstracts: Society of Magnetic Resonance in Medicine. Berkeley, Calif: Society of Magnetic Resonance in Medicine; 1993:328. Abstract.