| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, University of California at San Francisco, San
Francisco, CA.
2 Present address: Department of Radiology, Institut für
Röntgendiagnostik, Technische Universität München, Ismaninger
Str. 22, 81675 München, Germany.
3 UCSF & UCB Joint Graduate Group in Bioengineering, Berkeley, CA.
Received February 19, 2007;
accepted after revision March 26, 2007.
J. S. Bauer and S. Banerjee equally contributed to this study.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. MRI of the ankle was performed in three fresh human cadaver specimens and three healthy volunteers. Axial and sagittal T1-weighted, axial fat-saturated T2-weighted, and coronal intermediate-weighted fast spin-echo sequences, as well as a fat-saturated spoiled gradient-echo sequence, were acquired at 1.5 and 3 T. At 3 T, reduced data sets were reconstructed using a generalized autocalibrating partially parallel acquisition (GRAPPA) technique, with a scan time reduction of approximately 44%. All images were assessed by two radiologists independently concerning image quality. The signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) were measured in every data set. In the cadaver specimens, macroscopic findings after dissection served as a reference for the pathologic evaluation.
RESULTS. SNR and CNR in the GRAPPA images were comparable to the standard acquisition at 3 T. The image quality was rated significantly higher at 3 T with both normal and parallel acquisition compared with 1.5 T. There was no significant difference in ligament and cartilage visualization or in image quality between standard and GRAPPA reconstruction at 3 T. Ankle abnormalities were better seen at 3 T than at 1.5 T for both normal and parallel acquisitions.
CONCLUSION. Using higher field strength combined with parallel technique, MR images of the ankle were obtained with excellent diagnostic quality and a scan time reduction of about 44%. In addition, parallel imaging can provide more flexibility in protocol design.
Keywords: ankle pathology • cartilage pathology • GRAPPA • high-field 3-T MRI • ligament pathology • parallel imaging • scan time reduction • tendon pathology
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In contrast, parallel imaging reduces scanning time not by changing either the k-space bandwidth or the trajectory but by increasing the sampling interval along the phase-encoding axis. When the MR signal is received by an array of coils, the scan can be accelerated by acquiring fewer phase-encoding data points, and the missing data can be synthesized after the acquisition using the spatial-encoding information of all the coil elements, which is the underlying principle of parallel imaging. Because scanning time is directly proportional to the number of phase encodes, R-folds undersampling in the phase-encoding direction reduces scanning time by the same factor, although a penalty is paid in terms of the SNR [3].
Especially in standard clinical musculoskeletal protocols, scanning time reduction using parallel imaging can prevent motion artifacts and allows more flexibility in protocol design. Several different reconstruction algorithms have been proposed for parallel imaging, such as SENSE (sensitivity encoding) [3]; SMASH (simultaneous acquisition of spatial harmonics) [4]; and, more recently, mSENSE (modified SENSE) and GRAPPA (generalized autocalibrating partially parallel acquisition) [5].
Studies that focus on the application of parallel imaging in clinical musculoskeletal MRI at 1.5 T have been reported in the literature [6–9]. Among those, GRAPPA proved to be a reliable technique with the highest effective SNR while being robust to artifacts. However, a paucity of data exists regarding parallel imaging at 3 T [10, 11] and parallel imaging for small-FOV, high-spatial-resolution MRI [12]. Thus, for this study, we focused on small-FOV imaging of the ankle at 3 T with an autocalibrating parallel technique, and we compared the quality of the resulting images with that of standard acquisitions at 3 and 1.5 T. The comparison was performed in terms of SNR; CNR; image quality; ligament, tendon, and cartilage visualization; and assessment of ankle abnormalities.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Imaging
All imaging procedures were performed on a 1.5-T scanner and a 3-T scanner
(Signa, GE Healthcare). Both MR systems were equipped with gradients, having a
maximum gradient strength of 4 x 10–4 T/cm. For imaging
of the specimens, a four-channel receiver coil (NMSC-001, Nova Medical) was
used; for imaging of the volunteers, an eight-channel receiver coil (SHc-Head,
Medical Devices) was used. A standard ankle protocol that had been previously
optimized at our institution for an acquisition time of approximately 6
minutes for each sequence at 1.5 and 3 T was applied.
The protocol consisted of axial and sagittal T1-weighted, axial fat-saturated T2-weighted, and coronal fat-saturated intermediate-weighted fast spin-echo sequences and a fat-saturated spoiled gradient-recalled acquisition in the steady state (SPGR) for dedicated cartilage imaging (Table 1). In addition, at 3 T, parallel imaging acquisitions were obtained for 3D SPGR sequences and simulated for fast spin-echo sequences by decimating fully acquired data sets. The parallel acquisitions had to be simulated for the fast spin-echo sequences because the they were implemented only for gradient-echo sequences in the scanner at the time that this study was performed.
|
Partial parallel imaging was used with a reduction factor (R) of 2 and 16 calibration lines. Thus, the parallel imaging data sets consisted of 144 phase encodes compared with 256 in a conventional acquisition. This difference resulted in an acquisition time reduction of approximately 44% for the parallel scans compared with the acquisition time for the conventional scans at 1.5 T and at 3 T. The scanning times of the conventional acquisitions at 1.5 and 3 T were in the same range, as shown in Table 1.
All images were reconstructed offline on a workstation (Sun Workstation, Solaris) using reconstruction software programmed in Matlab (MathWorks). For the partially parallel imaging data sets, images were reconstructed using a GRAPPA-based algorithm [12, 13], whereas full imaging data sets were reconstructed using a standard Fourier reconstruction.
Image Analysis
SNR was measured in every sequence, and CNR was calculated for the SPGR
sequences. Images obtained from parallel imaging acquisition have a spatially
varying distribution of noise. In those images, SNR calculations, which are
based on mean signal intensity in a region of interest (ROI) and the
background noise, are not valid. Therefore, SNR and CNR were calculated from
twice-repeated acquisitions. An ROI was placed in the identical location on
each of the two images, and the average SNR in the ROI was determined as the
ratio of the average signal in the ROI in the mean image and the SD of the
signal in the difference image
[14]:
 |
All the images were assessed by two radiologists independently concerning image quality and visualization of anatomic structures and abnormalities. The radiologists were blinded to all information about the specimens and volunteers and to the field strength and type of reconstruction.
For all sequences, image quality was graded by each of the radiologists using a 4-level confidence scale. Items subjectively assessed with this scale included sharpness of edges; amount of blurring and noise; delineation of small ligamentous structures; and contrast between different tissues, such as joint fluid, cartilage, cortical bone, ligaments, and fat. Images were given a score of 4 if subjective image quality was optimal so that ligaments and cartilage were clearly demarcated and noise and artifacts were minimal to the eye. A score of 3 was given when one or two of the previously mentioned criteria were visually less than optimal. A score of 2 corresponded to pronounced limitations in the images that affected diagnostic quality. Images scored as 1 had substantial diagnostic limitations because of extensive artifacts and noise.
|
|
|
In the cadaver specimens, abnormalities were evaluated on 1.5- and 3-T scans on and parallel scans obtained at 3 T separately. Macroscopic findings after dissection served as the standard of reference. Ligament and tendon abnormalities were assessed using the fast spin-echo sequences. Although no ligament abnormalities were present, abnormalities of eight tendons (tibialis anterior and posterior, flexor and extensor hallucis longus, flexor and extensor digitorum longus, and peroneus brevis and longus) were described on a 5-point scale. The absence of abnormalities was scored as 0; tenosynovitis, which appear as pronounced fluid around the tendon, was scored as 1; tendinosis, which showed elevated, abnormal signal, as 2; partial split or full split tear as 3; and complete rupture as 4.
Cartilage abnormalities were evaluated on the SPGR sequences. The depth of a lesion was categorized using a modified Noyes classification [15]: 1, lesion < 50% of the cartilage thickness; 2, lesion > 50% of the cartilage thickness, 3, full-thickness lesion. Similar classification systems were used for macroscopic assessments.
Statistically significant differences in image quality, SNR, and CNR were tested using a paired two-sided Student's t test with a significance level of p < 0.05. All statistical computations were processed using SPSS software (version 11.5, SPSS).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
|
|
|
|
|
|
|
The image quality of all sequences was rated highest at 3 T (Table 3) for both specimen and volunteer studies. The axial T1-weighted sequence obtained an average score of 3.5 of 4 points at 3 T for both parallel and normal acquisitions, compared with a score of 2.8 at 1.5 T. There was no significant difference between parallel and normal acquisitions at 3 T for all sequences (p > 0.05), whereas 3 T performed significantly better than 1.5 T in three of five sequences (p < 0.05).
|
Similar trends were found for ligament visualization (Table 4). For all ligaments, visualization was rated highest for images obtained using a normal acquisition at 3 T (average score, 3.6). The average scores were slightly lower for parallel acquisition at 3 T (average score, 3.5), but this difference was not significant (p < 0.05). Both average scores were significantly higher than the average score at 1.5 T (p < 0.05); however, ligament visualization was good even at 1.5 T (average score, 3.3 of 4 points).
|
The pathologic examination of the three ankle specimens revealed three cartilage abnormalities that were classified as one grade 3 lesion in the lateral talus, one grade 2 lesion in the medial talus, and one grade 1 lesion in the tibia. Furthermore, in two specimens, three tendon abnormalities were found: a partial split tear of the peroneus tendon was seen in two specimens and a partial split tear associated with tenosynovitis of the peroneus longus tendon was seen in one specimen. In the third specimen, an os tibiale externum was present.
The tendon abnormalities were detected on all three sequence types. The grade 3 cartilage lesion was seen on images obtained using both parallel and standard acquisitions at 3 and 1.5 T. However, at 1.5 T, this cartilage lesion was described by only one of the two radiologists. The grade 1 and 2 cartilage lesions were missed in all three acquisitions by both radiologists.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Parallel imaging techniques such as GRAPPA are referred to as "autocalibrating" techniques because the coil calibration necessary for the parallel imaging reconstruction is built into the accelerated scan. Hence, the reconstruction is not affected by patient motion between scans or by gradient nonlinearity in the high-resolution scan. GRAPPA has several advantages over other parallel imaging techniques for our specific application. In small-FOV imaging, often, there is slight aliasing outside the ROI but within the FOV. Griswold et al. [5] showed that parallel imaging techniques that belong to the class of GRAPPA are more robust to such situations than SENSE-like parallel imaging techniques. Moreover, the quality of image reconstruction by the former class of parallel imaging techniques is not as sensitive to errors in estimation of coil sensitivities as the latter.
Figures 2B and 2E show axial GRAPPA-reconstructed images of the left ankle with slight aliasing from the right ankle. The aliasing in the FOV and difficulty of estimating the coil sensitivity in certain regions would compromise the image quality for a SENSE-like parallel imaging technique, whereas as can be seen in Figures 2B and 2E, the GRAPPA technique provided a good image quality.
The limitations for this preliminary study must be considered. First, coils optimized for musculoskeletal and parallel imaging were not available for both 3- and 1.5-T scanners. The benefit of parallel imaging is expected to be higher with optimized receiver arrays. Another potential limitation of this study is the small sample size and limited range and number of abnormalities that were analyzed. However, these factors do not affect the comparison of image quality as presented.
In conclusion, the combination of parallel and high-field MRI can provide high-spatial-resolution images with good diagnostic quality in a short scanning time. In addition to reducing scanning time, parallel imaging can provide more flexibility in protocol design—for example, with 3D sequences the technique can be used to improve volume coverage without increasing scanning time. Alternatively, if SNR permits, spatial resolution can be improved without increasing acquisition time compared with the standard acquisition. For 2D fast spin-echo sequences, a reduction of echo-train length will reduce T2 blurring and can improve the SNR compared with standard acquisitions. The reduction in the number of 180° pulses during the sequence will also decrease the radiofrequency power deposition in the subject, which can be a significant benefit at higher magnetic fields.
Although we used only a reduction factor of 2 for this study, higher reduction factors can be used with a receiver array optimized for musculoskeletal applications if the baseline SNR is sufficiently high. The results of this study showed that parallel imaging using the GRAPPA algorithm provides a robust way to further improve musculoskeletal imaging at high field strength. A large-scale clinical study evaluating different abnormalities in the ankle to determine the optimal combination of reduction factor, image quality, and time necessary for imaging each abnormality now is clearly warranted.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  D. K. Oh, Y. C. Yoon, J. W. Kwon, S.-H. Choi, J. Y. Jung, S. Bae, and J. Yoo Comparison of Indirect Isotropic MR Arthrography and Conventional MR Arthrography of Labral Lesions and Rotator Cuff Tears: A Prospective Study Am. J. Roentgenol., February 1, 2009; 192(2): 473 - 479. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J. Y. Jung, Y. C. Yoon, S.-H. Choi, J. W. Kwon, J. Yoo, and B.-K. Choe Three-dimensional Isotropic Shoulder MR Arthrography: Comparison with Two-dimensional MR Arthrography for the Diagnosis of Labral Lesions at 3.0 T Radiology, February 1, 2009; 250(2): 498 - 505. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. J. Stevens, R. F. Busse, E. Han, A. C. S. Brau, P. J. Beatty, C. F. Beaulieu, and G. E. Gold Ankle: Isotropic MR Imaging with 3D-FSE-Cube--Initial Experience in Healthy Volunteers Radiology, December 1, 2008; 249(3): 1026 - 1033. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
