| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology (Abdominal Imaging), University of Pittsburgh Medical
Center, Presbyterian Campus, 200 Lothrop St., Pittsburgh, PA 15213.
2 Instituto di Radiologia Universitaria, Azienda Ospedaliero Universitaria di
Udine, Udine, Italy.
Received July 11, 2007;
accepted after revision December 9, 2007.
Address correspondence to K. Hosseinzadeh
(hosseinzadehk{at}upmc.edu).
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Thirty-nine patients (18 men and 21 women; mean age, 51.8 years) were referred for MRCP. Multiangled thick-slab breath-hold SSFSE imaging was performed without and with integrated parallel imaging with sensitivity encoding (SENSE). Images were retrospectively reviewed. A 4-point rating scale was used to grade overall image quality and visibility of 10 ductal segments. A Wilcoxon's signed rank test compared the two techniques. Analysis of signal intensities and relative contrast of fluid-filled structures to background tissue on the basis of region of interest was calculated. Differences between the techniques were compared using a Student's t test.
RESULTS. Two-dimensional thick-slab SSFSE with integrated parallel imaging showed a mild but statistically significant increase in the relative contrast (p = 0.01) of ductal segments. Image quality deteriorated slightly with parallel imaging, but the results were not statistically significant (p = 0.06). Parallel imaging improved duct conspicuity of the medial and lateral segments of the left lobe and the posterior segment of the right lobe. However, statistical improvement was achieved only for the medial segment of the left lobe (p = 0.03). The mean scores of the remaining ducts were either unchanged or worse with parallel imaging, but the differences were not significant.
CONCLUSION. Although there was improvement in the relative contrast of bile to background tissue, improvement in overall image quality was not observed with parallel imaging. However, application of parallel imaging to thick-slab SSFSE may improve depiction of the smallest-caliber ducts.
Keywords: biliary tract • MR cholangiopancreatography • pancreatic ducts
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Parallel imaging techniques were first introduced to reduce scanning time, using the spatial sensitivity information inherent in an array of multiple receiver surface coils by reducing the number of time-consuming phase-encoding steps [19]. Parallel imaging techniques obtain a final full-field-of-view image with either a k-space algorithm, simultaneous acquisition of spatial harmonics (SMASH) and generalized autocalibrating partially parallel acquisition (GRAPPA), or an image-based algorithm, sensitivity-encoding (SENSE) technique [19-24].
A unique sequence-specific advantage of parallel imaging is the application to the SSFSE sequence. When parallel imaging is applied to a 2D thick-slab SSFSE sequence, the decrease in the number of phase-encoding steps may be used to reduce the length of the echo train, obtaining a shorter readout with less T2-related signal decay [25-27], thus improving image detail by reducing image blur. Consequently, reduction in image blur may translate into edge enhancement that would improve the definition of the central and peripheral ducts of the pancreaticobiliary system. In addition, SSFSE sequences can even gain signal-to-noise ratio (SNR) when parallel imaging is applied because late echoes are less affected by excessive T2 decay in the shortened parallel imaging-implemented echo train [25]. Ultimately, potential improvements in image quality may translate into improved diagnostic accuracy.
Although there have been previous investigations using parallel imaging in MRCP sequences [28-30], no study to our knowledge has been published in the radiology literature investigating the impact on image sharpness and detail when parallel imaging is applied to 2D thick-slab SSFSE MRCP acquisition. The purpose of this study was to quantitatively and qualitatively evaluate the effect of parallel imaging on image quality and anatomic detail in 2D thick-slab SSFSE MRCP images.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MRI Technique
MRI was performed on a 1.5-T superconducting system (Signa Excite HD, GE
Healthcare) with an 8-channel (12 elements) torso array coil after a fasting
period of at least 4 hours. Each patient underwent MRCP using a combination of
breath-hold or respiratory-triggered T2-weighted sequences. Breath-hold 2D
thick-slab SSFSE sequences without (TR range/TE, 2,507-3,528/900; echo-train
length, 222) and with parallel imaging (SENSE), acceleration factor of 2
(1,240-1,738/900; echotrain length, 103) were acquired on the same rotating
coronal oblique planes to display the biliary tree and pancreatic duct to
avoid overlapping fluid-filled structures. Other sequence parameters were kept
the same between these sequences (echo spacing, 9.5 milliseconds; bandwidth,
± 25.0 kHz; field of view, 300 mm; matrix size, 256 x 256; slab
thickness, 40 mm). The thick-slab images were prescribed from the transverse
SSFSE sequences, and a pause was used between thick-slab 2D SSFSE interleaved
acquisitions. Chemical-selective fat saturation was applied. Transmit and
receiver gains of both thick-slab 2D SSFSE sequences were kept constant to
ensure accurate quantitative evaluation during the data analysis. The TR was
automatically adjusted by the system with implementation of parallel imaging
as a consequence of the reduced echotrain length.
The other T2-weighted sequences included in the MRCP protocol consisted of breath-hold 2D multislice SSFSE (TR/TE, 1,499/90; field of view, 360 mm; matrix size, 256 x 256; slice thickness, 4 mm) acquired in the transverse and coronal planes, 3D fast recovery fast spin-echo (FRFSE) respiratory-triggered acquired in an oblique coronal plane and implemented with SENSE (1 x respiratory cycle/730; field of view, 320 mm; matrix size, 256 x 256; slice thickness, 1.4 mm; acceleration factor, 2). Finally, each patient underwent a T1-weighted, 2D spoiled gradient-echo (SPGR) sequence acquired in the transverse plane covering the pancreas with chemical-selective fat saturation (TR/TE, 185/3; flip angle, 70°; field of view, 360 x 250 mm; matrix size, 256 x 160; slice thickness, 8 mm). The retrospective data analysis was focused only on the 2D thick-slab SSFSE sequences acquired with and without SENSE implementation.
Quantitative Evaluation
The SNR in SENSE-implemented sequence is dependent on the acceleration and
geometric factors. The geometric factor varies over the image and describes
local noise enhancements to the final reconstructed image. In this way, the
background noise of parallel imaging varies with spatial position
[19,
26]. Therefore SNR
calculations, difficult to define in an unambiguous fashion, were not
performed in our study. Thus a single investigator performed the analysis of
the relative contrast between the fluid-filled bile and pancreatic ducts to
the background soft tissues for each of the 39 patients. The mean and SD of
signal intensities were measured on a PACS workstation (iSite, Philips Medical
Systems) by placing a circular region of interest (ROI) over the common bile
duct, the bifurcation of the common hepatic duct, the gallbladder, and the
pancreatic duct in the pancreatic head. The signal intensity of the background
was measured by placing an ROI over the parenchyma of the liver and the
pancreas, specifically over a region without signal from fluidfilled
structures, that is, between the left and right hepatic duct for the liver and
adjacent to the head of the pancreatic duct for the pancreas. The relative
contrast values for each ductal segment were calculated using the following
equation for both sequences
[31]:
 | (1) |
Qualitative Evaluation
Images from conventional and SENSE-implemented 2D thick-slab SSFSE
sequences for each of the 39 patients were independently and randomly reviewed
on a PACS workstation by two experienced abdominal radiologists with 10 and 3
years of experience. These reviewers were blinded to patient and sequence
information. To minimize any learning bias, images from each sequence were
reviewed separately in two reading sessions with a time interval of 1 week.
Also, each session was composed of a randomly generated mix of conventional
and parallel-imaging sequences for the 39 patients. An interval of at least 6
months between the acquisition of the images and the review sessions was
required.
Readers were unaware of the patient's clinical history, indications for the examination, and imaging reports. Each reader was requested to evaluate the overall image quality on the basis of visual evaluation of SNR and the presence of artifacts due to motion (respiration, pulsation) and aliasing. The evaluation was performed using a 4-point scale: 0 (nondiagnostic), severe artifacts or severe decrease of SNR precluding any evaluation of the pancreaticobiliary system; 1 (poor), moderate artifacts, low SNR, noisy image but the biliary tree and pancreatic duct are still recognizable partially or completely; 2 (good), minimal artifacts, good SNR, good evaluation of the biliary tree and pancreatic duct; and 3 (excellent), no artifacts, very good SNR, very good visualization of the biliary tree and pancreatic duct. Each reader assessed the conspicuity of 11 ductal segments: common bile duct and common hepatic duct (CBD/CHD); right hepatic duct; left hepatic duct; medial segment of the left hepatic duct; lateral segment of the left hepatic duct; anterior seg ment of the right hepatic duct; posterior segment of the right hepatic duct; third-order intrahepatic branches; and head, body, and tail of the pancreatic duct. Each segment, except the third-order branches, was graded on the basis of a 4-point scale: 0 (no visualization); 1 (poor visualization), visualization of only a part of the duct; 2 (moderate visualization), visualization of the whole duct until the next bifurcation with blurred margins; and 3 (complete visualization), visualization of the whole duct until the next bifurcation with sharp edges. A binomial scale evaluation was applied to the third-order branches: 0, no visualization; 1, visualization of at least a single third-order branch. A grade "NA" was assigned if the segment could not be evaluated either because of exclusion from the imaging volume or a nondiagnostic acquisition. Finally, differences in grading for both image quality and duct conspicuity were resolved by consensus.
Statistical Analysis
For the quantitative data, a Student's t test was used to evaluate
differences in relative signal intensities of the fluid-filled structures and
background (relative contrast) between the conventional and the
SENSE-implemented 2D thick-slab SSFSE sequences. For the qualitative data,
differences in image quality between the two sequences were compared using the
nonparametric Wilcoxon's test, which incorporated calculation of the grade
difference between the conventional and the SENSE-implemented sequences. In
addition, for each of the 10 duct segments, differences in duct conspicuity
between the conventional and the SENSE-implemented sequences were compared by
using the Wilcoxon's test. The level for statistical significance was set at
p < 0.05. All statistical computations were performed with
commercially available software (SPSS for Windows, version 13.0).
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
Qualitative Results
The overall image quality grades from the conventional and
SENSE-implemented 2D thick-slab SSFSE sequences obtained by consensus are
displayed in Table 2. Of the 39
patients considered, the majority of cases were graded of good quality: 23
(59.0%) cases with the conventional sequence and 25 (64.1%) cases with the
SENSE-implemented sequence. The SENSE-implemented sequence produced more
poor-quality images (five [12.8%] vs three [7.7%]) and fewer excellent quality
images than the corresponding conventional sequence (eight [20.5%] vs 12
[30.7%]) and in some cases these artifacts partially or completely obscured
the ducts (Figs. 1A and
1B). In only one patient
(2.6%), the overall image quality was graded nondiagnostic on both sequences.
Overall image quality of the conventional sequence was shown to be higher than
that of the corresponding SENSE-implemented sequence, with mean grades of 2.2
and 2.0, respectively, and a mean grade difference of 0.2. However, the
p value obtained with the Wilcoxon's test showed this difference not
to be statistically significant (p = 0.06).
|
|
|
The comparison of duct segment clarity between the two sequences was conducted using the consensus scores on 38 patients over the original population of 39. Ducts from one patient were excluded from the conspicuity analysis because both readers agreed in scoring each segment of the pancreaticobiliary system as NA. Table 3 displays the consensus results of the duct conspicuity scores on the basis of a segment-by-segment analysis. The 380 segment pairs evaluated using a 4-point scale system received an average score ranging from 1.9 to 2.8 when a conventional 2D thick-slab SSFSE sequence was used and from 1.5 to 2.8 when the corresponding SENSE-implemented sequence was used. Regarding the biliary ducts, SENSE-implemented 2D thick-slab SSFSE allowed better clarity of the medial and lateral segments of the left hepatic duct and the posterior segment of the right hepatic duct, with mean grade differences of -0.4, -0.1, and -0.2, respectively, whereas the right hepatic duct was best depicted using the conventional sequence, with a mean grade difference of 0.1. The CBD/CHD, left hepatic duct, and anterior segment of the right hepatic duct showed no difference in the conspicuity score. The bile third-order branches were visualized by both sequences in 35 patients (92.1%). Improvements in the SENSE-implemented sequence were mainly reflected in edge enhancement or sharpness of the duct margins (Figs. 2A and 2B). The head of the pancreatic duct received a higher score when evaluated by means of the SENSE-implemented 2D thick-slab SSFSE, with mean grade difference of -0.1. The tail of the pancreatic duct was best depicted by the conventional sequence, with a mean difference of 0.1, whereas no significant difference was noted between the two sequences in the conspicuity of the body of the pancreatic duct. The p value obtained with the Wilcoxon's test showed a statistically significant difference (p = 0.03) only for conspicuity of the medial segment of the left hepatic duct (Table 4 and Figs. 3A and 3B).
|
|
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
This technique can be performed either as a single thick-slab or multislice thin-section mode. When performed as a single thick-slab (20-70 mm) sequence, multiple acquisitions in variable coronal oblique planes are necessary to compensate for the lack of individual tomographic slices. However, the 2D thick-slab SSFSE mode offers a more efficient and faster visualization of the pancreaticobiliary ductal system than 3D reconstructed images obtained from a multislice acquisition technique [17]. The main drawback of 2D thick-slab SSFSE sequence is the long readout period, which results in strong T2 decay, reducing the amplitude of the high spatial frequency containing late echoes that are responsible for image sharpness. This attenuation may result in blurring along the phase-encoding direction and a decrease of image detail along with loss of signal intensity and image resolution [18, 25].
Parallel imaging was initially introduced [19] to reduce acquisition time, combining two methods for spatial encoding. Parallel imaging techniques use the spatial sensitivity information inherent in an array of multiple receiver surface coils to partially replace time-consuming spatial encoding usually performed by switching magnetic field gradients. Only a fraction of phase-encoding steps are directly acquired, resulting in accelerated image acquisition with unchanged spatial resolution [19-23]. The missing k-space lines can be obtained by means of k-space (SMASH, GRAPPA)- or image (SENSE)-based reconstruction algorithms [19-24]. SMASH and GRAPPA use the information from a fully sampled region of k-space (calibration scans) to calculate the weights for an interpolation filter that is then used to calculate the missing k-space data before the Fourier transformation [20, 21]. On the other hand, image-based reconstruction is performed after the aliased images are obtained from undersampled k-space. The SENSE algorithm creates an unaliased image by using the coil sensitivity maps to "unwrap" the aliasing artifacts.
Parallel imaging, when applied to the SSFSE sequence, offers a unique advantage in the reduction of the number of phase-encoding steps and therefore the length of the echo train and the duration of the readout period [25-27]. For a conventional sequence using a long echo-train length, T2 relaxation affects the transverse magnetization over the entire course of image acquisition and exacerbates the attenuation of the signal intensity in the echoes acquired later in the sequence that are responsible for image sharpness. This attenuation introduces blurring along the phase-encoding direction, with degradation of image detail and SNR [18]. A shorter echo-train length yields a shorter readout period and less signal attenuation because of the T2 decay. In this study, parallel imaging applied to SSFSE may improve image detail with a gain in spatial frequency information and, in contrast to other pulse sequences, may even provide a gain in SNR because the SNR loss associated with the reduced number of acquired k-space lines can be offset by the increase in peak signal intensity for the late echoes in the train (the T2-related signal decay is exponential) [25].
This study investigated the potential advantage of the implementation of parallel imaging when applied to SSFSE sequences. Two-dimensional thick-slab SSFSE MRCP sequences, obtained with and without parallel imaging (acceleration factor, 2), were compared qualitatively in terms of image quality and pancreaticobiliary system visualization and quantitatively as relative contrast. The quantitative evaluation in this investigation revealed a significant but small increase in relative contrast between fluid-filled structures and background tissues when SENSE was applied. The derivation of relative contrast as stated in equation 1 requires input of signal intensities of both fluid-filled structures and background tissues (liver and pancreas parenchyma). No difference in signal intensity of the fluid-filled structures was detected between the two sequences, whereas a reduction in background signal was noted with the implementation of parallel imaging. It is difficult to determine a single cause to explain the signal intensity differences using in vivo human data.
One potential explanation for the magnitude of the reduction in background tissue signal intensity would be to consider the relative T1 and T2 signal contribution from background tissues. During a SENSE-based SSFSE sequence, the default TR is automatically reduced to accommodate the lower echo-train length. In our protocol, the mean TRs for conventional and SENSE-implemented sequences were 3,000 and 1,500 milliseconds, respectively. Estimated T1 for background tissues (liver and pancreas) at 1.5 T is approximately 586 milliseconds [32]. Given the overlapping nature of the rotating thick-slab SSFSE sequence, there is potential for increased image saturation. Even though the acquisition includes a fixed pause between each TR interval to allow recovery of the longitudinal magnetization, a reduced TR would result in less recovery of the longitudinal magnetization before application of the next radiofrequency excitation pulse. Thus the signal becomes partially saturated. Therefore the application of parallel imaging will cause a reduction of the T1 component of the signal from the background tissue, translating to an increase in relative contrast.
The estimated T2 value for background tissues (liver and pancreas) at 1.5 T is approximately 42 milliseconds [32]. In determining the T2 contribution to the background signal, the length of the readout train will be a function of the interecho spacing and echotrain length. With the implementation of parallel imaging, the total duration of the echotrain approximates 1 second at a reduced echo-train length of 103 and fixed interecho spacing. Therefore the majority of the background tissue signal will have decayed, given its low T2 value, and the echoes acquired during the end of the echo train, which contribute to fine detail, will have negligible signal intensity. The T2 contribution of the background tissue signal is negligible and remains unchanged with parallel imaging.
A similar principle applies to fluid-filled structures except that the longer T2, on the order of seconds, may retain more signal with a shorter readout train and compensate for any T1 signal loss. It has been increasingly recognized in the literature that the inherent SNR loss of parallel imaging can often be outweighed by the reduction in T2-related signal loss that it affords for long readout sequences such as those used here. Because time constraints are not a factor during 2D thick-slab SSFSE acquisition, an increase in TR would allow a longer recovery interval between readouts in addition to the fixed pause, producing better fluid enhancement because more longitudinal recovery is available for the next readout train. Ideally a controlled experiment can be conducted using phantoms reflecting the different tissues and calculating the respective T1 and T2 values. Accurate signal intensity and SNR calculations can be made with application of parallel imaging techniques when coil sensitivity profiles are reproduced accurately.
The qualitative analysis showed that the conventional SSFSE 2D thick-slab sequence provides high-quality MRCP images without any significant improvement after SENSE implementation. Although a quantitative increase in relative contrast was shown when applying parallel imaging, no significant contribution was observed in the visual grading of image quality or duct conspicuity. Although not statistically significant, the SENSE-implemented sequence resulted in a slight decrease in overall image quality because of the introduction of artifacts and noise, often described with errors in the reconstruction process, which at times caused partial obscuration of the central ducts.
With regard to duct conspicuity, no improvement was shown with the implementation of parallel imaging other than the medial segment of the left hepatic duct, the duct receiving the lowest mean grade in our series. These results show that the potential increase in image sharpness and detail related to the reduction in echo-train length is not apparent for the larger fluid-filled ducts because of the inherent long T2 values exhibited by these bile-filled structures against relatively low background signal. On the other hand, the medial segment of the left hepatic duct is a small-caliber duct no greater than one or two pixels in size, and the improved conspicuity is likely a reflection of the benefits rendered from the shorter echo-train length leading to improved image detail and edge enhancement.
The image quality and duct conspicuity achieved with the 2D thick-slab SSFSE MRCP sequence is comparable with other reports in the literature [14-16]. Recently, some studies have appeared in the radiology literature investigating the role of parallel imaging when applied to MRCP sequences [28-30]. Of these studies, only one has investigated differences in image quality with implementation of parallel imaging to the same sequence. Asbach et al. [29] determined that the implementation of parallel imaging to a respiratory-triggered 3D FRFSE MRCP sequence does not significantly influence image quality or duct conspicuity, even with a distinct reduction in imaging time. In our study, time efficiency is not as critical for the 2D thick-slab SSFSE sequence because of the very short acquisition time. To our knowledge, this is the first study investigating the potential role of parallel imaging in improving 2D thick-slab SSFSE MRCP sequences in terms of visualization of the pancreaticobiliary system.
One limitation of the study was that the analysis was conducted retrospectively on a small number of patients. We compared image quality rather than the diagnostic accuracy of the 2D SSFSE thick-slab MRCP techniques for biliary and pancreatic abnormalities. A sufficient sample size for correlative imaging and intraoperative findings was not available, and a meaningful statistical evaluation would not be possible given the wide range of potential abnormalities present in this group. One might expect that the discriminatory threshold of the 2D thick-slab SSFSE sequences being compared would be less critical in ductal diseases in which there are dilated segments. The impact of a change in spatial resolution on image quality was not explored.
A comparison with the 3D FRFSE respiratory-triggered sequence was not made as this sequence intrinsically shows higher contrast and SNR, which is in keeping with the nature of the pulse sequence—namely, a lower receiver bandwidth, signal averaging, and volumetric excitation. The 2D SSFSE thick-slab sequence shows excellent inplane spatial resolution, eliminates respiratory motion artifacts secondary to an erratic breathing pattern, and is part of the routine MRCP protocol.
A further minor limitation of our study was that the implementation of parallel imaging to the 2D thick-slab SSFSE may have benefited from additional optimization. Given the short breath-hold times for thick-slab 2D SSFSE, the lower default TR setting recommended by the vendor may not be optimal with implementation of parallel imaging, and increasing TR to equate that of the conventional sequence will allow greater recovery of the longitudinal magnetization and limit image saturation. As discussed earlier, parallel imaging may be used to improve image matrix and spatial resolution; however, the corresponding decrease in SNR would then need to be addressed.
In summary, we evaluated image quality and anatomic detail by implementing parallel imaging to 2D thick-slab SSFSE MRCP. Although qualitative assessment revealed more noise and artifacts in the parallel imaging-implemented sequence, there was no significant difference in overall image quality when compared with the conventional sequence. On the other hand, there was minimal improvement in the relative fluid-filled structures to background contrast with parallel imaging; however, these were not sufficient for observable qualitative differences. Despite the theoretic benefits of parallel imaging in the SSFSE sequence, there is realistic statistical equivalence of the two techniques for duct conspicuity. Further parameter optimization may be warranted to translate the relative contrast enhancement to overall improvement in image quality and duct conspicuity.
Acknowledgments
We are grateful to Fernando Boada for his assistance on the technical
component of this manuscript and to Howard E. Rockette for the statistical
analysis.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
