| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 MR Centre for MS Research, Vrije Universiteit Medical Centre, P. O. Box 7057,
1007 MB Amsterdam, The Netherlands.
2 Department of Clinical Epidemiology and Biostatistics, Vrije Universiteit
Medical Centre, Van der Boechortstr. 7, 1081 BT Amsterdam, The
Netherlands.
Received December 20, 2001;
accepted after revision February 21, 2002.
Address correspondence to F. Barkhof.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Serial MR images were obtained at baseline, month 6 (n = 20), and month 7 (n = 16). For the half-yearly intervals, subtracted 3D FLAIR images and T2-weighted spin-echo images were compared. For the monthly intervals, subtracted 3D FLAIR images were compared with triple-dose contrast-enhanced T1-weighted spin-echo images. New, enlarging, and enhancing lesions were marked in consensus by two radiologists.
RESULTS. At the half-yearly intervals, 3D FLAIR imaging detected more new or enlarging lesions than T2-weighted spin-echo imaging, both at the initial interpretation (80 vs 52; p < 0.001) and after a side-by-side comparison of the lesions (88 vs 65; p < 0.001). Post hoc analyses showed the largest benefit for new (rather than enlarging), for small, and for temporal lesions. At the monthly intervals, 32 enhancing lesions were detected on contrast-enhanced T1-weighted spin-echo images versus 20 new or enlarging lesions detected on 3D FLAIR images (p < 0.05). After a side-by-side comparison of the lesions, seven additional lesions were identified on 3D FLAIR images, making the difference with contrast-enhanced T1-weighted spin-echo images insignificant (27 vs 32; p > 0.05).
CONCLUSION. Isotropic 3D FLAIR imaging holds great promise for the detection of new or enlarging lesions in multiple sclerosis using registration and subtraction techniques certainly at longer intervals.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Registration and subtraction of images could be further improved by the use of isotropic voxels. Recently, the use of multislab three-dimensional (3D) fast fluid-attenuated inversion recovery (FLAIR) imaging with isotropic voxel dimensions for the detection of lesions in patients with multiple sclerosis was reported [11]. In our current study, we investigate the use of serial 3D FLAIR images with small isotropic voxels for the detection of new or enlarging lesions on subtracted images. To this end, 3D FLAIR images were compared with subtracted conventional two-dimensional (2D) T2-weighted spin-echo images for the detection of new and enlarging lesions at half-yearly intervals. Further, the sensitivity of the 3D FLAIR sequence in the detection of new and enlarging lesions seen at monthly intervals was evaluated in comparison with contrast-enhancement on T1-weighted spin-echo imaging.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
At baseline and at month 6, dual-echo 2D T2-weighted spin-echo images (proton density—weighted and T2-weighted) and 3D FLAIR images were obtained. Two-dimensional T2-weighted spin-echo imaging and 3D FLAIR imaging were performed alternatively as a first sequence to avoid any bias. At month 7, the imaging protocol consisted of 3D FLAIR imaging followed by 2D T1-weighted spin-echo imaging. After the injection of 0.3 mmol/kg of gadopentetate dimeglumine via a long IV line (to prevent change in head position), contrast-enhanced 2D T1-weighted spin-echo images were obtained after a delay of 15 min.
MR Imaging
MR imaging was performed on a 1.5-T scanner (Vision; Siemens, Erlangen,
Germany). To minimize motion of the head, we used a vacuum cushion to
stabilize the head in the standard circularly polarized head coil. The
isotropic 3D FLAIR sequence was performed as two interleaved axial series,
each consisting of six 12-mm thick slabs (with 10 partitions each) with the
following parameters: TR/TE, 6500/120; inversion time, 2200 msec; matrix, 162
x 256; field of view, 196 x 310 mm; total acquisition time, 21 min
[11]. It is based on a turbo
spin-echo imaging sequence, using a turbofactor of 27. Imaging parameters of
the T2-weighted spin-echo sequence were 2600/20; excitations, 80; matrix, 256
x 256; field of view, 250 x 250; and total acquisition time, 16
min. For the T1-weighted spin-echo sequence, the imaging parameters were
800/15; excitations, 1; matrix, 256 x 256; field of view, 250 x
250; acquisition time, 10 min. To obtain contiguous 3-mm-thick slices, we
combined two interleaved sets of 23 images (1-mm in-plane resolution) with a
3-mm gap.
Image Postprocessing
Because of the imperfect slab profiles, a venetian blind effect was
observed in 3D FLAIR images in the slice-encoding direction, which was
corrected as described previously
[11]. For image realignment,
we used an automatic voxel-based registration algorithm on the basis of the
mutual information similarity measure
[13,
14] as a matching criterion.
In the mutual information theory, one considers the measure of dispersion in
the distribution of the gray values of an image, the entropy of an image: a
contrast-rich image has a high entropy value, whereas a homogeneous image has
a low value. If two images are registered (i.e., their mutual information is
maximal), the joint entropy of the overlapping part of two images is low.
Trilinear interpolation was used for both image interpolation and reslicing of
data. The 3D FLAIR images and the 2D T2-weighted spin-echo images obtained at
month 6 were registered to baseline, and the subtracted images were created by
subtracting baseline from month 6. The 3D FLAIR images obtained at month 7
were registered to month 6, and month 6 images were subtracted from month
7.
The interobserver agreement for the detection of enhancing lesions in multiple sclerosis has been shown to be good using original images [15]. Because it is not to be expected that registration and subtraction techniques will add a substantial improvement in contrast to the detection of new or enlarging T2 lesions [9, 10], image registration and subtraction were not performed for T1-weighted spin-echo images.
Image Analysis
Two radiologists with experience in the analysis of registered and
subtracted images documented new or enlarging lesions on the subtracted images
in consensus, with the original and registered images as a reference to
confirm that a change was genuine.
For the 2D T2-weighted spin-echo images, the subtracted proton density—weighted images were used in the analysis of new or enlarging lesions. Contrast-enhanced 2D T1-weighted spin-echo images were compared with the unenhanced images to identify enhancing lesions.
After identification of the lesions, all images were reviewed side-by-side,
and lesions were compared and classified according to size (small diameter,
5 mm; large diameter, > 5 mm) and location: periventricular and
nonperiventricular in the frontal, parietal, or occipital regions and the
temporal, juxta-cortical, or infratentorial regions. The following
classification for false-negative, false-positive, and retrospectively
identified lesions was used. If a lesion was seen on one sequence, and in
retrospect also on the original (nonsubtracted) images of another sequence,
the lesion was classified as false-negative for the latter sequence. When a
lesion was identified as active on one sequence, whereas it was not seen on
the other because it was already present on the first scan, the lesion was
classified as false-positive. If a lesion was, in retrospect, seen as new or
enlarging during the comparison of images, we classified it as such.
The McNemar test was used for comparison of the number of lesions after the initial interpretation and after the side-by-side comparison of each sequence and for the post hoc comparisons in the subgroup analysis.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
On subtracted images, a new lesion can be identified as a bright area against a gray background (Fig. 1A,1B,1C), whereas an enlarging lesion can be identified as a bright area adjacent to a preexisting lesion at baseline. Slight reslicing artifacts may be seen on registered images, especially at the brain surface and gray and white matter borders, resulting in artifacts on subtracted images. Inconstant occurrence of pulsation artifacts and flow in vessels may also lead to artifacts on subtracted images. For instance, such an area could be hypointense on the baseline 3D FLAIR image but isointense on the follow-up image, resulting in a bright area on the subtracted image (Fig. 2A,2B,2C). The authenticity of new and enlarging lesions on the subtracted images was therefore always confirmed by comparing the original and registered images.
|
|
|
|
|
|
Detection of Lesions at Half-Yearly Intervals: 3D FLAIR Versus 2D
T2-Weighted Spin-Echo
Between baseline and month 6, three patients had not developed either new
or enlarging lesions on any sequence. After the initial interpretation, 3D
FLAIR imaging depicted 80 lesions (58 new and 22 enlarging) compared with 52
(39 new and 13 enlarging) for the 2D T2-weighted spin-echo imaging (p
< 0.001). A side-by-side comparison revealed a considerable number of new
and enlarging lesions, in retrospect, more for the 2D T2-weighted spin-echo
images than for the 3D FLAIR images. An extra five new and two enlarging
lesions were seen on 3D FLAIR imaging, whereas on T2-weighted spin-echo
imaging, six lesions were considered new (three of which were juxtacortical)
and six were enlarging, in retrospect. Only a few lesions were reclassified as
false-negative or false-positive. Because of subtraction artifacts, each
sequence failed to reveal a small infratentorial lesion. Further, one large
temporal lesion was not detected on T2-weighted spin-echo imaging because of
subtraction artifacts. One small nonperiventricular lesion was classified as
false-positive on a 2D T2-weighted spin-echo image because this lesion was
apparent on both baseline and 6-month 3D FLAIR images and therefore not
visible on the subtracted images. After a final review of lesions, we found
that 3D FLAIR imaging still performed better than 2D T2-weighted spin-echo
imaging in the depiction of new or enlarging lesions (88 vs 65; p
< 0.001). Figure
3A,3B,3C,3D,3E,3F
illustrates a new juxtacortical lesion on subtracted 3D FLAIR imaging that was
not seen on 2D T2-weighted spin-echo imaging.
|
|
|
|
|
|
Post hoc analysis showed that the number of enlarging lesions was not significantly different (p = 0.424), but that the difference between 3D FLAIR imaging and 2D T2-weighted spin-echo imaging was mostly accounted for by new lesions (p = 0.003). Three-dimensional FLAIR imaging detected more small and large lesions than 2D T2-weighted spin-echo imaging, but only the number of small lesions was significant, with 52 versus 36 lesions (p < 0.005). When the difference in the number of lesions was compared according to location, the detection of temporal lesions (small and large combined) was most benefited by the use of 3D FLAIR imaging, with 25 versus 11 lesions (p = 0.001).
Detection of Lesions at Monthly Intervals: 3D FLAIR Versus
Contrast-Enhanced 2D T1-Weighted Spin-Echo
Between month 6 and month 7, nine of the 16 patients with multiple
sclerosis did not show an enhancing, new, or enlarging lesion. After the
initial interpretation, we detected 43 enhancing lesions on the
contrast-enhanced 2D T1-weighted spin-echo images. However, 11 enhancing
lesions were not seen as either new or enlarging on the subtracted 3D FLAIR
images because the lesions were already present at month 6; these lesions were
classified as false-positive. Figure
4A,4B,4C,4D
shows such a lesion in the posterior fossa that was already present at month 6
on the 3D FLAIR image and diminished in size at month 7, resulting in a
hypointense area on the substracted image. With the exception of these 11
lesions (which most likely were already enhancing at month 6 or even before
and reflect persistent but not new enhancing lesions), contrast-enhanced 2D
T1-weighted spin-echo imaging detected 32 new enhancing lesions, whereas 3D
FLAIR imaging detected 20 new or enlarging lesions (p = 0.017).
|
|
|
|
In retrospect, seven enhancing lesions were identified on 3D FLAIR imaging: four as new (three of which were small) and three as enlarging lesions. Five new enhancing lesions appeared to be enlarging on 3D FLAIR imaging. Four new lesions (three temporal and one new nonperiventricular) and one enlarging nonperiventricular lesion were depicted only on 3D FLAIR imaging. Ten small enhancing lesions could not be identified as either new or enlarging on 3D FLAIR imaging. After a side-by-side comparison of lesions, the number of lesions revealed on 3D FLAIR imaging was 27 compared with 32 enhancing lesions revealed on contrast-enhanced T1-weighted spin-echo imaging (p > 0.05).
Post hoc analyses showed that contrast-enhanced T1-weighted spin-echo MR imaging depicted more small lesions than 3D FLAIR (20 vs 12, p = 0.04), whereas the latter detected (slightly) more large lesions (12 vs 15, p > 0.05). No statistically significant difference was found among the number of lesions according to location (p > 0.05).
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The use of subtraction to detect the change of disease burden in multiple sclerosis (using nonisotropic voxels) appears to be efficient and reliable [10]. Because such procedures are most efficient when applied to images with small isotropic voxels, we compared 3D FLAIR images with the typically used 3-mm 2D T2-weighted spin-echo images. The significantly higher number of small new lesions depicted on 3D FLAIR imaging is not surprising. The 3D sequence has a high contrast-to-noise ratio and the additional advantage of being more suitable for registration and subtraction because of the use of (high-resolution) isotropic voxels. The clinical significance of the additional lesions revealed on 3D FLAIR imaging should be evaluated in a larger clinical study.
At month 7, we used triple-dose contrast-enhanced T1-weighted spin-echo images because such a contrast dose combined with delayed imaging was shown to be a very sensitive method to detect enhancing lesions in multiple sclerosis [23]. Approximately 25% of the enhancing lesions were not seen as either new or enlarging on subtracted 3D FLAIR images because they were already present at month 6. These lesions most likely would have been enhancing lesions at month 6. Of the remaining 32 enhancing lesions, 3D FLAIR images detected 47% after the initial interpretation and 68% after the side-by-side comparison of lesions. The sensitivity of 3D FLAIR imaging to detect active lesions is further supported by the observation that six enhancing lesions were seen as enlarging lesions on 3D FLAIR images and thus already present at month 6 and that five new or enlarging lesions depicted on 3D FLAIR images did not show enhancement on triple-dose contrast-enhanced T1-weighted spin-echo images. The nature of these lesions depicted on 3D FLAIR images but not depicted on contrast-enhanced T1-weighted spin-echo images remains unclear; perhaps these lesions would enhance on an image obtained at month 8 or would have been enhancing for less than a month between month 6 and month 7. Further studies are needed to confirm these findings and should include monthly enhanced scans. Conversely, 10 small enhancing lesions were not seen as either new or enlarging on 3D FLAIR images. This finding indicates that T1-weighted spin-echo imaging and 3D FLAIR imaging are complementary methods in the detection of disease activity in patients with multiple sclerosis.
False-negative scores due to postprocessing artifacts from the registration or subtraction procedures occurred in only three (1.7%) of 180 lesions, of which two were small lesions. This finding indicates that the combination of the mutual information similarity measure as a matching criterion and trilinear interpolation is suitable for the detection of new or enlarging lesions in multiple sclerosis using registered and subtracted images. However, care should be taken not to identify a bright area as a lesion without confirmation from the original (registered) images. This procedure can be performed quickly and reliably by simultaneously scrolling through the matched original registered and subtracted images. No difficulty should be expected in the differentiation of a hypointense lesion and an artificially hypointense area on the original FLAIR images [24] because the latter has the same form as the structure introducing the flow artifact but is displaced along the phase-encoding direction. On nonsubtracted FLAIR images, genuinely hypointense lesions are surrounded by a bright rim, whereas flow artifacts are not.
The long acquisition time of 21 min for the 3D FLAIR sequence is a potential drawback for patients with multiple sclerosis and results from the use of an interleaved multislab implementation of the 3D FLAIR technique. In the ideal case, the entire brain is imaged as one single-slab volume acquisition with a short acquisition time instead of interleaved multislabs [25, 26], but such a 3D FLAIR sequence with isotropic 1-mm voxels has yet to be developed.
In conclusion, serial 3D FLAIR imaging using isotropic 1.2-mm voxels seems to hold great promise in the detection of lesions in patients with multiple sclerosis. Our data indicate that 3D FLAIR imaging performed at half-yearly intervals in the detection of new or enlarging multiple sclerosis lesions is superior to 2D 3-mm T2-weighted spin-echo MR imaging. We also found that 3D FLAIR imaging performed at monthly intervals in the detection of multiple sclerosis lesions is equivalent to triple-dose contrast-enhanced T1-weighted spin-echo MR imaging, although the latter sequence seems to provide complementary information at short intervals.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  Y. Duan, P.G. Hildenbrand, M.P. Sampat, D.F. Tate, I. Csapo, B. Moraal, R. Bakshi, F. Barkhof, D.S. Meier, and C.R.G. Guttmann Segmentation of Subtraction Images for the Measurement of Lesion Change in Multiple Sclerosis AJNR Am. J. Neuroradiol., February 1, 2008; 29(2): 340 - 346. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. J. G. Geurts, P. J. W. Pouwels, B. M. J. Uitdehaag, C. H. Polman, F. Barkhof, and J. A. Castelijns Intracortical Lesions in Multiple Sclerosis: Improved Detection with 3D Double Inversion-Recovery MR Imaging Radiology, July 1, 2005; 236(1): 254 - 260. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. P. van der Voorn, P. J. W. Pouwels, W. Kamphorst, J. M. Powers, M. Lammens, F. Barkhof, and M. S. van der Knaap Histopathologic Correlates of Radial Stripes on MR Images in Lysosomal Storage Disorders AJNR Am. J. Neuroradiol., March 1, 2005; 26(3): 442 - 446. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. J. G. Geurts, L. Bo, P. J. W. Pouwels, J. A. Castelijns, C. H. Polman, and F. Barkhof Cortical Lesions in Multiple Sclerosis: Combined Postmortem MR Imaging and Histopathology AJNR Am. J. Neuroradiol., March 1, 2005; 26(3): 572 - 577. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Abstracts Multiple Sclerosis, July 1, 2003; 9(4_suppl): S1 - S153. [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
