| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Developments |
1 From the Department of Radiology, University of Virginia Health System, East Hospital, Rm 1063, Lee St, Box 800170, Charlottesville, VA 22908. Received October 25, 2000; revision requested December 15; revision received March 7, 2001; accepted March 29. Supported in part by grant NS35142 from the National Institute of Neurological Disorders and Stroke and by Siemens Medical Systems. Address correspondence to D.F.K. (e-mail: dfk3b@virginia.edu).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
Index terms: Artifact • Brain, MR, 10.12149 • Cerebrospinal fluid, 167.12149 • Magnetic resonance (MR), pulse sequences, 10.12149
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
The promise of high sensitivity in the detection of SAS lesions with FLAIR has been mitigated by suboptimal specificity resulting from CSF-related artifacts with high signal intensity. Virtually all investigators who evaluated the accuracy of FLAIR imaging have noted these high-signal-intensity artifacts, which are present within both the SAS and ventricles (1,6,7). These artifacts, which are most pronounced in the posterior fossa, almost certainly result from the inflow of CSF that was originally outside the region affected by the inversion pulse. Although several investigators have attempted to diminish these artifacts by increasing the width of the inversion slab, none were successful in their eradication (1,7,8). Image quality in the posterior fossa is often further degraded by pulsatile motion of nonsuppressed CSF or blood in the major vessels, which results in artifactual signal intensities overlying the anatomy of interest.
Our group has developed a three-dimensional (3D), single-slab, FLAIR pulse sequence with high spatial resolution. Because this 3D pulse sequence uses nonspatially selective radio-frequency pulses, including the inversion pulse that affects the entire imaging slab, we hypothesized that high-signal-intensity artifacts from inflow of noninverted CSF, as well as artifacts from pulsatile motion, would be diminished or eradicated. The purpose of our study was to test this hypothesis.
|
Materials and Methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
The images were printed on film for review. Filming was performed by an experienced technologist, who chose window and level parameters appropriate for FLAIR imaging in routine clinical practice.
Image Interpretation
Four readers interpreted the images. Three readers had subspecialty training in neuroradiology, and one was a radiology resident. Images were interpreted independently and in random order. One of the subspecialty-trained neuroradiologists (D.F.K.) was aware of the study design but remained blinded to sequence type. The other three readers, to avoid bias, were told only that the pulse sequences were intended to yield dark CSF; they were blinded to pulse sequence type, hypothesis of the study, number of pulse sequences being compared, number of study subjects, and all identifying subject information. With a five-point graded response scale, the readers rated the degree of high-signal-intensity CSF artifacts in the fourth ventricle and at four locations within the SAS (upper cervical, prepontine, perimesencephalic, and suprasellar cisterns) as follows: grade 1, diffuse high signal intensity; grade 2, majority (>50%) of cistern filled with high signal intensity; grade 3, minority (<50%) of cistern filled with high signal intensity; grade 4, equivocal areas of high signal intensity; and grade 5, absence of high signal intensity. In addition, readers identified individual cranial nerves where possible. Last, the presence and severity of artifacts from vessels and/or CSF pulsation were graded by using an analogous scale (1 = worst, 5 = best). The ratings of high-signal-intensity CSF artifacts and pulsation artifacts were compared between the 2D and 3D images by using Wilcoxon rank-sum analysis. The numbers of identifiable cranial nerves were also compared between the two groups by using similar statistical tests. Interreader variability was evaluated by using a 
statistic, where agreement was defined as readings between two readers within one category on the five-point graded response curve. The degree of agreement was considered to be almost perfect (0.8–1.0), substantial (0.6–0.79), moderate (0.4–0.59), fair (0.2–0.39), slight (0–0.19), or poor (<0) (11).
|
Results |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
|
|
|
|
|
|
Cranial Nerve Identification
The mean numbers of cranial nerves confidently identified were 1.3 and 3.8 for 2D and 3D imaging, respectively (P = .006) (Figs 1, 2). The third, fifth, and VII/VIII complex were the most frequently identified cranial nerves.
Interreader Variability
statistics ranged from 0.67 to 0.74 between various readers, which is indicative of substantial interreader agreement (11).
|
Discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
More generally, the single-slab 3D FLAIR technique offers the advantages commonly associated with true 3D acquisition: thin, contiguous sections that result in reduced partial volume averaging of small structures compared with the thicker sections typically used in 2D acquisitions; the capability, with appropriate acquisition parameters, to obtain high-spatial-resolution images in arbitrary planes; and potentially higher signal-to-noise performance per unit time than that with a comparable 2D pulse sequence.
Other investigators have attempted modifications to the 2D FLAIR pulse sequence in hopes of reducing CSF-related artifacts. The most common technique reported includes widening of the inversion region, on the order of 50% thicker than the imaging section, to diminish inflow of noninverted magnetization (1,7,8). The use of wider inversion regions, however, does not necessarily eradicate the artifact. Bakshi et al (1), using a 2D FLAIR pulse sequence optimized for lesion conspicuity, imaging time, and artifact suppression, reported CSF-related artifacts with high signal intensity in the fourth ventricle in 50% of subjects. Conversely, in our study, high signal intensity within the fourth ventricle was absent in all subjects with our single-slab 3D FLAIR pulse sequence. Note that the use of a wider inversion pulse is, in effect, similar to the application of spatial presaturation slabs, which are commonly used for suppressing the signal from blood flow, on both sides of the image section. The usefulness of this approach is inherently limited by the effect of the wide inversion pulse on adjacent image sections.
Tanaka et al (12) recently reported on the use of a nonspatially selective inversion pulse with 2D FLAIR. With this approach, the inversion time is different for each section. To address this issue, all sections were acquired twice, once in ascending order and once in descending order, and then averaged to achieve approximately the same contrast properties for each section. Because a nonselective inversion pulse is used, the degree of artifact suppression should be comparable to that with our single-slab 3D technique. With the 2D technique, however, the image contrast varies somewhat between sections. In addition, the total time over which sections can be acquired following each inversion pulse must be limited to maintain acceptable contrast in the averaged images. This results in a relatively low anatomic coverage per unit time; in the study by Tanaka et al (12), only 11 sections were acquired in 4.5 minutes.
Other researchers have investigated the use of a multislab 3D FLAIR pulse sequence for central nervous system imaging (2,13,14). Similar to our single-slab technique, the multislab implementation offers the advantage, compared with 2D techniques, of thin contiguous sections with a clinically acceptable signal-to-noise ratio. Nonetheless, substantial artifacts may still occur in regions of large CSF pulsation such as the posterior fossa (14). Furthermore, multislab imaging possesses several important limitations compared with a single-slab approach. Because of section profile effects, some of the outer sections in each slab are usually discarded, thus decreasing the efficiency. Power deposition is relatively high and may compromise the coverage obtained per unit time, particularly at fields of 1.5 T and above. Each slab also experiences unwanted magnetization-transfer effects from the large number of high-flip-angle off-resonance radio-frequency pulses applied to the other slabs during the acquisition (15).
Suppression of the CSF signal on T2-weighted images may also be achieved without the use of the inversion recovery–based approach of FLAIR. Essig et al (16) recently demonstrated a technique in which CSF-suppressed images are obtained by subtracting an image with a very long echo time from an image with a moderate echo time; both images were obtained within the same acquisition by using a fast spin-echo pulse sequence. Because an inversion pulse is not used, this technique may provide increased contrast for certain lesions, such as those in multiple sclerosis, for which relatively poor conspicuity in the posterior fossa has been reported with FLAIR imaging (17,18). Nonetheless, to our knowledge, an investigation of the artifact characteristics for this dual-echo technique in the posterior fossa has yet to be presented. Thus, we cannot judge how it is likely to compare with our 3D approach for those processes that are well seen with FLAIR imaging.
Our series, although useful for demonstrating the potential benefits of single-slab 3D FLAIR imaging for suppressing high-signal-intensity CSF and pulsation artifacts, has several shortcomings. Only a small number of healthy volunteers underwent imaging. Even the small number of subjects, however, was sufficient for demonstrating a statistically significant reduction in artifacts with 3D versus 2D imaging. In addition, our volunteers were relatively young, and intraventricular artifact may increase with increasing age as CSF flow rates increase. Relatively severe artifacts, however, were seen even in the young volunteers in our series with 2D imaging, and these artifacts were markedly reduced with 3D imaging.
The 8-minute acquisition time used with our single-slab 3D FLAIR technique is several minutes longer than that for many pulse sequences routinely used for clinical MR imaging at 1.5 T. Thus, although our method appears to provide excellent suppression of high-signal-intensity CSF and pulsation artifacts, patient movement and subsequent image degradation during the relatively long imaging time is a concern. With our current pulse sequence, we use nonspatially selective radio-frequency pulses for inversion, excitation, and refocusing of the magnetization. These nonselective pulses, combined with radio-frequency transmission with the body coil, which is standard for head imaging with our MR system, resulted in aliasing (wraparound artifact) in the section direction unless the imaging slab covered the complete sensitive volume of the head coil. Therefore, for this study, the total imaging slab thickness was approximately 50% greater than that required to cover the anatomy of interest. We are developing an implementation of 3D FLAIR that uses a selective excitation radio-frequency pulse, which would enable the slab thickness to be matched to the volume of interest. This would permit the acquisition time to be decreased from 8 to approximately 5 minutes. Nonselective pulses would still be used for inversion and refocusing, thereby maintaining artifact-suppression and sampling-efficiency characteristics equivalent to those of our current technique.
Another potential criticism of our study is that we acquired 5-mm-thick sections with the 2D pulse sequence and 3-mm-thick sections with the 3D technique. The 5-mm section thickness used for the 2D pulse sequence was chosen to match that routinely used at our institution in the clinical environment. The 3-mm section thickness chosen for the 3D technique represented a tradeoff between shortening the acquisition time, which favors thicker sections, and limiting truncation artifacts in the section direction to an acceptable level, which favors thinner sections. In 3D acquisitions, truncation artifacts in the section direction become pronounced when the section thickness approaches the characteristic dimension of structures along that direction (19). We acknowledge that the thinner sections used for 3D versus 2D images biased our cranial nerve evaluation in favor of the 3D pulse sequence, although the capability for obtaining thin, contiguous sections with clinically adequate signal-to-noise ratios is an advantage for the 3D method.
Finally, although we have shown that the single-slab 3D FLAIR technique is useful for reducing artifacts, we have not demonstrated its ability to reveal abnormalities in the SAS or ventricles. It is possible that the 3D FLAIR pulse sequence may be less sensitive to these abnormalities than is the 2D FLAIR technique. Further studies are required to address this issue.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Author contributions: Guarantor of integrity of entire study, D.F.K.; study concepts, D.F.K., J.P.M.; study design, D.F.K.; literature research, D.F.K.; clinical studies, D.F.K., F.K.H.; data acquisition, D.F.K., F.K.H.; data analysis/interpretation, D.F.K.; statistical analysis, D.F.K., F.K.H.; manuscript preparation, D.F.K., J.P.M.; manuscript definition of intellectual content, D.F.K.; manuscript editing, J.P.M.; manuscript revision/review, D.F.K., J.P.M.; manuscript final version approval, D.F.K.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and Methods Results DiscussionREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  B. Moraal, D. S. Meier, P. A. Poppe, J. J. G. Geurts, H. Vrenken, W. M. A. Jonker, D. L. Knol, R. A. van Schijndel, P. J. W. Pouwels, C. Pohl, et al. Subtraction MR Images in a Multiple Sclerosis Multicenter Clinical Trial Setting Radiology, February 1, 2009; 250(2): 506 - 514. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Lisanti, C. Carlin, K. P. Banks, and D. Wang Normal MRI Appearance and Motion-Related Phenomena of CSF Am. J. Roentgenol., March 1, 2007; 188(3): 716 - 725. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. J. W. Pouwels, J. P. A. Kuijer, J. P. Mugler III, C. R. G. Guttmann, and F. Barkhof Human Gray Matter: Feasibility of Single-Slab 3D Double Inversion-Recovery High-Spatial-Resolution MR Imaging Radiology, December 1, 2006; 241(3): 873 - 879. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Cianfoni, M.G.M. Martin, J. Du, J.R. Hesselink, S.G. Imbesi, W.G. Bradley, and G.M. Bydder Artifact simulating subarachnoid and intraventricular hemorrhage on single-shot, fast spin-echo fluid-attenuated inversion recovery images caused by head movement: A trap for the unwary. AJNR Am. J. Neuroradiol., April 1, 2006; 27(4): 843 - 849. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I L. Tan, R. A. van Schijndel, P. J. W. Pouwels, H. J. Ader, and F. Barkhof Serial Isotropic Three-Dimensional Fast FLAIR Imaging: Using Image Registration and Subtraction to Reveal Active Multiple Sclerosis Lesions Am. J. Roentgenol., September 1, 2002; 179(3): 777 - 782. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
