| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Technical Innovation |
1 Department of Radiology, University of Wisconsin School of Medicine, 600
Highland Ave., CSC E1/374, Madison, WI 53792.
2 GE Healthcare, Waukesha, WI.
3 Global Applied Science Laboratory, GE Healthcare, Madison, WI.
4 Departments of Medical Physics, Biomedical Engineering, and Medicine,
University of Wisconsin, Madison, WI.
Received July 22, 2006;
accepted after revision June 10, 2007.
The employment status of J. McGovern and J. H. Brittain at GE Healthcare
did not influence the data in this study.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CONCLUSION. Compared with fat-saturated SPGR, IDEAL–SPGR imaging combined with parallel imaging at 3 T provides robust fat–water separation and significant improvement in cartilage SNR. Use of IDEAL–SPGR also led to dramatic improvement in cartilage–fluid contrast-to-noise ratio compared with fat-saturated SPGR imaging. Thus, use of IDEAL–SPGR may improve the accuracy of cartilage volume measurements and detection of cartilage surface defects. Excellent evaluation of the morphologic features of the knee cartilage with high-resolution, high-SNR images can be performed in 5 minutes.
Keywords: cartilage • fat suppression • knee • MRI • spoiled gradient echo • water–fat separation
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A variety of MRI sequences have been developed for cartilage imaging. These
sequences can be broadly divided into physiologic and morphologic techniques.
Physiologic techniques, which are used to detect biochemical abnormalities
before development of a morphologic abnormality, include T2
shortening [3],
23Na-MRI to detect regions of glycosaminoglycan depletion
[4], diffusion-weighted imaging
[5], T1
imaging
[6], and delayed contrast
enhancement [7]. Most
physiologic methods are investigational.
Clinical evaluation of cartilage has primarily focused on morphologic evaluation. Detection of the subtle early changes of osteoarthritis, that is, surface cartilage defects such as fissures, requires images of high resolution and high signal-to-noise ratio (SNR) with good fat suppression and good contrast between cartilage and synovial fluid. To be clinically practical, imaging must be completed in a reasonable time. Accurate quantification of cartilage volume is critical for clinical trials of drug therapies for osteoarthritis. Quantification enables detection of small changes in cartilage thickness and volume that would not be evident clinically or radiographically. Accurate volume measurements, which rely on computer segmentation [8], require good fat suppression, high SNR and, most important, excellent contrast between joint fluid and cartilage. This contrast is optimal when synovial fluid signal intensity is low compared with the relatively high signal intensity of cartilage.
In clinical practice, the most commonly used sequences for cartilage evaluation are 2D fat-saturated fast spin-echo and 3D spoiled gradient echo (SPGR) with fat saturation [9, 10]. Because of its high resolution, 3D coverage, and high cartilage signal intensity, fat-saturated SPGR has been considered the clinical reference standard for morphologic quantification and volumetric assessment of cartilage [9]. The intermediate signal intensity of synovial fluid on images obtained with fat-saturated SPGR, however, can limit sensitivity to small lesions, especially those on the articular surface of the cartilage, and can make it difficult to perform accurate computer segmentation to measure cartilage thickness and volume.
The major limitation of fat-saturated SPGR that has prevented widespread clinical acceptance is the prolonged imaging time, up to 10–12 minutes [9]. Such long imaging times are needed for fat-saturated SPGR because the sequence is SNR-limited and requires multiple averages to improve SNR performance. The poor SNR performance is closely related to the SNR inefficiency of the SPGR pulse sequence and is worsened by the high overhead of playing a fat-saturation pulse every TR.
Iterative decomposition of water and fat with echo asymmetry and least-squares estimation (IDEAL) [11] is used for robust separation of fat and water with very high SNR efficiency. Unlike conventional fat-saturation methods, IDEAL is insensitive to magnetic field (B0 and B1) inhomogeneity. Although acquisition of three separate images with slightly different TEs is needed to decompose water from fat, IDEAL is highly SNR-efficient, with an effective averaging value of 3.0 [12, 13]. This effective averaging is equivalent to the SNR performance that would be obtained if the object were to contain only water or fat and the three images were averaged together. In addition, IDEAL does not require the overhead of a fat-saturation pulse, and in comparison with fat-saturated SPGR, it is highly efficient [10, 11]. For these reasons, IDEAL–SPGR is more SNR-efficient than fat-saturated SPGR and can lead to a substantially shorter TR [14, 15]. Finally, fat-saturation methods such as fat-saturated SPGR may suffer from reduced SNR as a result of partial water saturation from fat-saturation pulses.
The need for three signal averages for IDEAL–SPGR can lead to prolonged imaging times. By taking advantage of the spatial information inherent in phased-array radiofrequency coils, parallel imaging can be used to reduce this acquisition time. The purposes of this study were to combine IDEAL–SPGR with parallel imaging at 3 T to acquire high-resolution, high-SNR images of cartilage across the entire knee within 5 minutes, an acceptable imaging time for a sequence added to routine clinical protocols, and to compare this new sequence with the traditional fat-saturated SPGR technique.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Parameters for both fat-saturated SPGR and IDEAL–SPGR imaging included a 512 x 224 matrix (zero-filled to 512 x 512), 16-cm2 field of view, 1.0-mm slice thickness, 90 slices, and acquisition bandwidth of ± 41 kHz. The resulting voxel dimensions were 0.3 x 0.7 x 1.0 mm. Linear autoshimming was used for both sequences. Optimal flip angles to maximize cartilage signal intensity were calculated on the basis of the Ernst angle with use of the TR for each sequence and a presumed cartilage T1 signal time of 1,200 milliseconds at 3 T [16, 17].
IDEAL–SPGR imaging was performed with the minimum possible TR of 10.8
milliseconds. The TEs were selected on the basis of previous findings
[13] that showed optimal SNR
performance when the phase between water and fat for the center echo is 
/2
+ k (k is any integer), that is, when water and fat
signals are in quadrature. The two remaining echoes are placed before and
after the center echo by a water–fat phase shift of 2/3 relative to
the center echo. This combination of water–fat phase shifts optimizes
the noise performance of IDEAL for decomposition of water from fat signals.
Several echo groups are possible, each with a different integer value of
k. To maximize the pulse sequence efficiency and achieve the shortest
TE (full echo), we chose the group in which the shortest echo equaled or
exceeded the minimum TE determined with other sequence parameters. The TEs for
this study corresponded to a k value of 4 based on a fat–water
chemical shift of –420 Hz at 3 T. This value corresponds to TE values of
4.6, 5.4, and 6.1 milliseconds. At 3 T, the spacing between echoes is 0.80
milliseconds, and the spacing between subsequent k groups is 1.2
milliseconds. To measure SNR when parallel imaging is used, IDEAL–SPGR
acquisition was performed twice in succession on each knee (see later).
We used a sensitivity encoding–based [18] parallel acceleration method (ASSET, GE Healthcare) with an acceleration factor of 2.22 for the IDEAL–SPGR acquisition. The optimized flip angle for imaging cartilage with IDEAL–SPGR was 8° based on a TR of 10.8 milliseconds, and the total imaging time was 4 minutes 59 seconds. Separate fat, water, and recombined images were reconstructed with an online reconstruction algorithm. Recombined images were corrected for chemical shift artifact in the readout direction by realigning the separated water and fat images by the chemical shift, which depends on imaging parameters, and is easily calculated (Yu H et al., presented at the 2004 annual meeting of the International Society for Magnetic Resonance in Medicine). Although this implementation of IDEAL involved three excitations, the absence of the fat-saturation pulse made the total imaging time less than three times greater than that of fat-saturated SPGR with a single excitation. Therefore, an acceleration factor of 2.22 resulted in equal imaging times for IDEAL–SPGR with three excitations and fat-saturated SPGR with a single excitation.
Fat-saturated SPGR imaging was performed with the minimum possible TR of 14.8 milliseconds, TE of 4.6 milliseconds (full echo), and a 9° flip angle, which is the optimized flip angle for imaging cartilage with fat-saturated SPGR based on a TR of 14.8 milliseconds. We used one average and no parallel imaging for a total imaging time of 5 minutes.
For SNR measurement, regions of interest (ROIs) were defined by a board-certified radiologist with 6 years of experience in the patellar cartilage, femoral condyle cartilage, muscle, and synovial fluid. The ROI shape was ovoid and optimized for best sampling of the tissue of interest. Although ROI placement was customized for each knee, the same ROI was used for all imaging sequences in each knee. The ROI size averaged 24 mm2 (114 pixels) for patellar cartilage and 8 mm2 (38 pixels) for joint fluid.
In conventional SNR measurement techniques, it is assumed that the distribution of noise across the imaged field of view is uniform [19, 20]. Because noise is not uniform across images reconstructed with parallel imaging methods [18], conventional methods for measuring SNR cannot be used. Therefore, the SNR measurements for IDEAL–SPGR were made with the difference method [21] on the basis of two identical acquisitions. SNR for fat-saturated SPGR images was measured with conventional techniques with one ROI on the image and the noise ROI on a region outside the body. SNR was calculated as the quotient of the average signal from a small ROI of each studied tissue and the SD of the noise in an ROI placed on a region outside the body and free of artifact. Because images were reconstructed with the square root of sum of squares technique [22] for multiple coils, a correction factor of 0.70 was used to adjust for the fact that noise in the magnitude image has a chi-square distribution that causes underestimation of noise [20].
To validate comparison between SNR measurements obtained with the two techniques, two identical fat-saturated SPGR acquisitions were obtained in four knees. SNR measurements were performed with both techniques with identical ROIs in patellar and femoral cartilage, synovial fluid, and muscle. These results were compared by use of a two-tailed Student's t test.
For each knee, cartilage–fluid contrast-to-noise ratio (CNR) was calculated as the difference between the SNR of the cartilage and the SNR of fluid. SNR and CNR values were compared by use of a paired Student's t test. Statistical comparisons were considered significant at p < 0.05. All statistical calculations for SNR measurements were made with Microsoft Excel 2004 version 11.0 for Apple Macintosh.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
SNR results for the eight knees are plotted in Figures 2A and 2B. The SNR of patellar cartilage was 34% higher for IDEAL–SPGR (20.2 ± 4.0) than for fat-saturated SPGR (15.0 ± 2.8) (p < 0.001). The SNR of femoral trochlear cartilage was 31% higher with IDEAL–SPGR (22.5 ± 3.4) than with fat-saturated SPGR (17.2 ± 1.6) (p < 0.01). Muscle SNR was 26% higher for IDEAL–SPGR (16.2 ± 3.6) than for fat-saturated SPGR (12.9 ± 2.3) (p < 0.001). For the SNR of synovial fluid, no significant difference was found between IDEAL–SPGR (12.5 ± 2.0) and fat-saturated SPGR (12.2 ± 1.8) (p < 0.001).
|
|
|
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Because of the absence of a fat-saturation pulse, an acceleration factor of 2.22 allows three effective signal averages of IDEAL–SPGR in the same acquisition time as a single average of fat-saturated SPGR. Because both acceleration factor and number of signal averages are related to SNR by the square root of imaging time, we would expect an SNR advantage of the square root of 3/2.22, or 1.17, 17% higher for IDEAL–SPGR than for fat-saturated SPGR, assuming optimal coil geometry and ignoring the g-factor. This value compares to an observed SNR advantage of 34% for patellar cartilage in this study. Additional factors that may have led to this discrepancy between expected and observed SNRs include partial water saturation caused by the fat-saturation pulse and inadvertent miscalibration of the radiofrequency transmit pulse. The result would be a flip angle different from the intended value and optimized for the Ernst angle on the basis of published values of the T1 signal of cartilage. Compared with fat-saturated SPGR, the IDEAL–SPGR technique showed a 2.7 times greater patellar cartilage–fluid CNR. This improvement in CNR addressed a major limitation of fat-saturated SPGR for morphologic evaluation of cartilage. The absence of the fat-saturation pulse allows a substantially shorter TR for IDEAL–SPGR (10.8 milliseconds) compared with the minimum TR obtainable with fat-saturated SPGR (14.8 milliseconds). This decrease in TR likely explains the improved T1 contrast between fluid and cartilage. Synovial fluid is better suppressed with shorter TR because of the long T1 signal of fluid.
The receiver bandwidth was chosen to minimize echo and TR. Although this relatively high bandwidth also limits the SNR, the same value was used for both sequences studied and therefore did not affect the comparison of SNRs of the two.
The marked improvement in CNR may improve the sensitivity for detection of cartilage surface defects with IDEAL–SPGR compared with fat-saturated SPGR. It is also an important advantage for automated segmentation and volume measurements [8, 23, 24]. Such quantitative measurements are critical for assessment of promising drug therapies aimed at slowing, stopping, and reversing the cartilage loss of osteoarthritis.
When it was compared with conventional SNR measurement in four knees imaged with two identical fat-saturated SPGR acquisitions, the difference method was shown to underestimate SNR by approximately 5% for all tissue types. No significant difference was seen when only cartilage was included in the comparison. The slight underestimation of SNR with the difference method can likely be attributed to misregistration between the two sequential acquisitions owing to minimal knee movement. Any misregistration increases the variance of the subtracted image and simulates noise, leading to underestimation of SNR. As might be expected, this effect seemed more evident in muscle, which is more heterogeneous, than in cartilage, which is more homogeneous. The small underestimation of SNR with the difference method suggests that the true SNR benefit of IDEAL–SPGR over fat-saturated SPGR may actually be slightly greater than these results indicate.
With a change from 1.5 to 3 T comes a doubling in chemical shift, from –210 to –420 Hz. This effect can be a liability for many sequences at 3 T, necessitating doubling band-width to reduce the increased chemical shift artifact and sacrificing much of the SNR benefit of higher field strengths.
Imaging at 3 T improves the performance of the IDEAL method, resulting in closer spacing of echoes and of consecutive echo groups, further improving the SNR efficiency of the technique. The reduced spacing between echo groups greatly improves the timing flexibility of the pulse sequence [13]. In addition, the ability of IDEAL to correct for chemical shift artifact obviates the need to increase bandwidth when increasing field strength to 3 T to reduce chemical shift artifact, allowing realization of the full SNR benefit of imaging at 3 T.
Although variants of IDEAL have been developed that include all three echoes within a single TR (Reeder SB et al., presented at the 2006 annual meeting of the International Society for Magnetic Resonance in Medicine), most implementations of IDEAL require three excitations with one echo per TR. This requirement can lead to prolonged imaging times but provides excellent SNR. Parallel imaging methods are highly complementary to IDEAL. Parallel imaging reduces the imaging-time penalty of IDEAL, and the high SNR performance of IDEAL offsets the SNR penalty of parallel imaging. As our findings show, combining IDEAL with parallel imaging can provide an optimal combination of SNR and reasonable imaging time.
In addition to fat-suppressed water-only imaging, IDEAL yields separate fat, in-phase, and out-of-phase images with no additional imaging time penalty. As an adjunct to routine clinical sequences, the images may contribute to increased confidence in the diagnosis of noncartilaginous abnormalities, such as bone marrow edema, and of ligamentous and meniscal injuries. Recent work [25] has shown the value of in-phase and out-of-phase imaging for characterization of benign versus malignant disease in vertebral bone marrow. Such images, which can be acquired with the IDEAL sequence, may aid in evaluation of incidental extremity marrow findings. The recombined images are corrected for chemical shift artifact in the readout direction, allowing use of lower receiver bandwidth with an associated increase in SNR.
Finally, the combination of high-resolution, high SNR, and improved fluid-cartilage CNR of IDEAL offers potential for improved segmentation and volumetric analysis studies in addition to its benefits for clinical evaluation of cartilage morphology [8, 23, 24]. Future studies may prove the predicted increased accuracy of segmentation, as was the case in a study in which porcine knee specimens (with saline displacement techniques as a reference standard) were used to evaluate the advantage of 3 T over 1.5 T for fat-saturated SPGR [26].
In summary, this study showed that the combination of parallel imaging with IDEAL for SPGR imaging of knee cartilage provides high-quality, high-SNR images of the cartilage with better SNR performance than achieved with fat-saturated SPGR and a greatly improved cartilage–fluid CNR.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
