| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Radiology, Memorial Sloan-Kettering Cancer Center, 1275 York
Ave., Rm. C278, New York, NY 10021.
2 Department of Diagnostic Radiology, Yale University School of Medicine, New
Haven, CT 06520.
3 Department of Radiology, New York University Medical Center, New York, NY
10021.
4 Department of Diagnostic Imaging, Valley Health System, Ridgewood, NJ
07450.
Received June 16, 2005;
accepted after revision October 21, 2005.
Address correspondence to J. Zhang
(zhangj12{at}mskcc.org).
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Thirty consecutively registered patients (14 women, 16 men; average age, 60.2 years; age range, 32-87 years) underwent imaging at 1.5 T with a 6-element body array coil. All imaging was performed with three MRCP techniques: free-breathing 3D T2-weighted TSE (TR/TE, 1,300/680; flip angle, 180°; field of view, 250-300 mm; matrix size, 256 x 256; slice thickness, 1 mm; parallel acquisition technique factor, 2); breath-hold 3D T2-weighted TSE (same parameters as the free-breathing 3D technique); breath-hold coronal and oblique coronal thick-slab 2D TSE without parallel acquisition technique (2,800/1,100; flip angle, 150-180°). Quantitative measures of image signal and contrast were evaluated by analysis of variance and paired Student's t tests. A 5-point scale (1, nondiagnostic, to 5, high diagnostic confidence) was used to compare the 3D and 2D data sets for image quality and definition of biliary and pancreatic ductal anatomic features. Friedman's nonparametric and Wilcoxon's rank sum tests were performed for statistical analysis of the qualitative assessments.
RESULTS. Quantitative results showed free-breathing and breath-hold 3D TSE images had significantly higher relative signal intensity and contrast than 2D TSE images (p < 0.0001). The qualitative findings showed that both free-breathing and breath-hold 3D TSE techniques gave better delineation of biliary anatomy (p < 0.0001) than the 2D technique. The overall quality of 3D images was better than that of 2D images, and 3D imaging was better at depicting pancreatic ducts, although the difference did not reach statistical significance.
CONCLUSION. Three-dimensional volumetric MRCP images are of superior quality and give better delineation of pancreaticobiliary anatomy than conventional 2D images and have the added advantage of multiplanar and postprocessing capabilities.
Keywords: cholangiopancreatography • MRI
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Current MRCP techniques typically entail use of thin-section single-shot fast spin-echo sequences and thick-section heavily T2-weighted sequences to produce images of the biliary tree and pancreatic duct [3-8]. Two-dimensional thick-slab acquisition usually consists of a single T2-weighted image (usually 30-80 mm in slice thickness) acquired in the coronal and coronal-oblique planes. Although this method produces images that appear similar to ERCP images, the technique has intrinsic limitations. Because of the large section thickness and associated volume averaging, subtle pathologic features such as a small intraductal filling defect due to calculus or tumor can be obscured by surrounding hyperintense fluid signal [2]. In addition, given the limited number of projection views, complicated anatomic details often cannot be delineated. Therefore, for diagnosis, most centers rely primarily on additional thin-slice 2D single-shot fast spin-echo acquisitions through the biliary system. These sequences typically have a slice thickness of 4-6 mm and because they are half-Fourier methods, they typically have relatively limited spatial resolution, signal-to-noise ratio (SNR), and contrast-to-noise ratio.
The potential advantages of 3D T2-weighted MRCP techniques over 2D imaging include the capacity for use of thinner sections without interslice gaps and a higher SNR and for postprocessing manipulation of the images with techniques such as multiplanar reconstruction, maximum intensity projection, and volume rendering. With these tools, unlimited projectional views in any arbitrary plane can be obtained to clarify anatomic relationships. To date, several groups [9-15] have reported that 3D T2-weighted turbo spin-echo (TSE) MRCP techniques have diagnostic accuracy [9-12, 14] similar to that of conventional 2D techniques. Most of the 3D approaches, however, have been hindered by long acquisition times and constraints on anatomic coverage and spatial resolution. Acquisition times for most published methods have ranged from 8 to 12 minutes with respiratory triggering to produce images with section thicknesses of 2-3 mm [9-14].
With technical advances in MRI, including faster gradients and parallel acquisition technique, marked improvement in spatial resolution can be achieved in shorter acquisition times. The purpose of this study was to compare the image quality of a 3D heavily T2-weighted TSE MRCP sequence with approximately 1 x 1 x 1 mm voxel size with the image quality of conventional 2D heavily T2-weighted sequences. We assessed two versions of the 3D method, one with acquisition times short enough for a single breath-hold and a second performed with respiratory triggering.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Imaging was performed with a 1.5-T system (Symphony with Quantum gradients or Sonata, Siemens Medical Solutions) with maximum gradient strength of 30-40 mT/m, slew rate of 125-200 T/m/s, and a 6-element body array coil placed anteriorly and a spinal phased-array coil placed posteriorly. Informed consent was obtained from all patients before imaging, and the study was approved by the institutional review board.
Each patient underwent imaging with three MRCP techniques: respiration-triggered 3D heavily T2-weighted parallel TSE imaging with the patient instructed to breathe normally; breath-hold 3D heavily T2-weighted parallel TSE imaging; and breath-hold conventional 2D coronal and coronal-oblique thick-slab heavily T2-weighted TSE imaging. To null fluid signal in the stomach and duodenum, each patient drank two cups of pineapple juice 5-10 minutes before the examination unless a contraindication, such as NPO (nothing by mouth) status or history of allergy to pineapple, was present or the patient could not tolerate drinking the juice [16].
MR Sequences and Parameters
Breath-hold 2D T2-weighted TSE thick-slab sequence—Images
were acquired in the coronal and two coronal oblique planes (
±
30° to the coronal plane so that at least one acquisition was in the plane
of the pancreatic duct). A 2D heavily T2-weighted TSE sequence with the
following parameters was performed: TR/TE, 2,800/1,100; refocusing flip angle,
150-180°; bandwidth, 130 Hz/pixel; slice thickness, 40-80 mm; rectangular
field of view, 250-300 mm; matrix size, 256 x 256; typical voxel size,
1 x 1 x 40-80 mm; acquisition time, 2 seconds for each
image.
Breath-hold 3D T2-weighted TSE sequence with parallel acquisition
technique—Images were acquired in the coronal plane. A 3D heavily
T2-weighted two-shot TSE sequence with the following parameters was performed:
1,300/680; refocusing flip angle, 180°; bandwidth, 255 Hz/pixel;
rectangular field of view, 250-300 mm; matrix size, 256 x 256;
echo-train length, 120 depending on rectangular field of view; zero
filling; partial-Fourier, 6/8 in slice selection direction; partitions, 26-30
with 2-mm thickness and zero filling into 52-60 slices; interpolation into a
typical voxel size of 1 x 1 x 1 mm; parallel acquisition technique
factor, 2; acquisition time, 28-30 seconds; generalized autocalibrating
partially parallel acquisition with 24 reference lines for parallel imaging. A
-90° radiofrequency pulse (restore pulse) at the end of the echo train was
applied to flip the transverse magnetization back to the longitudinal
direction to shorten spin relaxation time.
3D T2-weighted TSE sequence with respiratory triggering—The foregoing sequence was performed but with respiratory triggering with the patient breathing freely. All acquired parameters were otherwise the same as for the breath-hold technique, and the voxel size was typically 1 x 1 x 1 mm with a parallel acquisition technique factor of 2. Respiratory triggering was performed with prospective navigator technique whereby after each trigger, one echo-train length of approximately 120 (depending on rectangular field of view) was obtained. To acquire 26-30 partitions, which were zero filled to generate 52-60 partitions, a total of 26-30 respiratory cycles were needed. Acquisition times were typically 1.5-3 minutes.
Data Analysis: Quantitative Assessment of Image Quality
Quantitative and qualitative assessments of the image quality of each
technique were performed on a commercially available MR workstation (Syngo,
Siemens Medical Solutions). Blinded to imaging results, a single investigator
performed region-of-interest analysis for each of the three sets of images for
each of the 30 patients. The mean and SD of signal intensity were determined
for the common bile duct and liver (Fig.
1A,
1B). Care was taken to place
regions of interest within normal-appearing portions of each organ, away from
intrahepatic bile ducts. Relative signal intensity and relative contrast
parameters were used for direct comparison of image quality for the
nonparallel acquisition technique (2D) and parallel acquisition technique (3D)
[17]. Relative signal
intensity (SI) of the common bile duct (CBD) was calculated as
SICBD/SDCBD. Relative contrast of common bile duct to
liver was calculated as (SICBD -
SILiver)/(SDCBD2 +
SDLiver2)1/2.
|
|
Image quality was based on scoring of overall image quality, delineation of biliary anatomic and pathologic features, and delineation of pancreatic ductal anatomic and pathologic features as detailed later. The following scale was used for evaluation of overall image quality: 1, unreadable; 2, poor quality; 3, satisfactory quality; 4, good quality; and 5, excellent quality. The diagnostic quality for delineation of biliary anatomic and pathologic features was assessed by scoring of a number of parameters.
|
|
|
The following scale was used for evaluation of clarity of pancreatic ducts: 1, unreadable; 2, extreme blur; 3, moderate blur; 4, mild blur; and 5, sharp. The confidence level in determining the presence of pancreatic ductal anomaly and pathologic changes was graded as: 1, nondiagnostic; 2, difficult to determine, low confidence; 3, some difficulty, moderate confidence; 4, mild difficulty, good confidence; and 5, little to no difficulty, high confidence.
All artifacts were documented, and the following scale was used: 1, unreadable study; 2, severe artifacts; 3, moderate artifacts; 4, mild artifacts; and 5, no artifacts. For determination of biliary and pancreatic anatomic and pathologic features, the combination of the results with the three sequences reached by consensus of the two reviewers was used as the reference standard.
Statistical Analysis
A separate univariate analysis was conducted for each of the quantitative
and qualitative measures of image quality. Specifically, the three imaging
techniques (free-breathing 3D T2-weighted TSE, breath-hold 3D T2-weighted TSE,
and T2-weighted thick slab-sequence) were analyzed in a pairwise comparison
with respect to relative signal intensity and relative contrast. Tukey's
honestly significant difference procedure was used in the context of two-way
analysis of variance and in terms of the consensus evaluations of each
qualitative measure of image quality performed with Friedman's test with a
Bonferroni correction to the significance level. In each case, patient
identification was treated as the blocking factor, and the analyses were
conducted with a familywise type 1 error rate of 5%. All statistical
computations were performed with SAS System for Windows software, version 9.0
(SAS Institute). All p values reported for the comparison of the
quantitative and qualitative assessment were subjected to Tukey's honestly
significant difference procedure and Bonferroni correction, respectively.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Quantitative Results
Mean (± SD) values for quantitative measurements of relative common
bile duct signal and relative common bile duct-to-liver contrast for the three
MRCP sequences are given in Table
1. The free-breathing 3D TSE and breath-hold 3D TSE images had
significantly greater relative signal intensity and contrast than the
thick-slab 2D TSE images (p < 0.0001). There was a trend that
free-breathing 3D TSE images had higher relative signal intensity and contrast
compared with breath-hold 3D TSE images, but this difference was not
statistically significant.
|
Qualitative Results
Mean (± SD) values for subjective measures of image quality are
shown in Table 1. Overall, the
qualitative values supported the quantitative results. Most images were good
to excellent in overall quality, with an average score of 4 or greater. Both
free-breathing and breath-hold 3D TSE techniques showed better delineation of
biliary anatomy (p < 0.0001) and cystic duct insertion (p
= 0.002) than the 2D technique. Although the differences did not achieve a
threshold of p < 0.05 for statistical significance, a consistent
trend was observed that both 3D techniques had higher qualitative scores than
the 2D thick-slab technique in the following aspects: overall image quality,
clarity of bile ducts and confidence in delineation of biliary pathologic
features, and clarity of pancreatic ducts and confidence in delineation of
pathologic features in the pancreas (Fig.
3A,
3B,
3C). In general, the images
obtained with the two 3D techniques were similar in quality except in
delineation of anatomic and pathologic features of pancreatic ducts (Fig.
4A,
4B,
4C), for which the 3D
free-breathing images scored slightly higher than the 3D breath-hold images.
The 3D free-breathing technique showed more artifacts compared with the other
two techniques, mainly because of blurring related to respiratory motion (Fig.
5A,
5B,
5C). High spatial resolution
and relative signal intensity and relative contrast, however, compensated for
the artifacts, which did not appear to compromise diagnostic quality.
|
|
|
|
|
|
|
|
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Experience with 3D T2-weighted MRCP methods has been limited in widespread implementation because of long acquisition times (as long as 11 minutes), constraints on anatomic coverage, and spatial resolution limited to 3-mm-thick sections [9-14]. Wielopolski et al. [15] described a 3D segmented echoplanar MRCP technique with high spatial resolution (voxel size, 1.8 x 1.0 x 1.0 mm) and breath-hold acquisition times that were close to those we report. Those authors reported good image quality and depiction of ducts at least 2 mm in diameter. With a smaller voxel size (1.0 x 1.0 x 1.0 mm), our TSE approach enabled visualization of ducts 1 mm in diameter. Whether the TSE approach is also less sensitive to susceptibility artifacts from bowel gas that might hinder echoplanar imaging remains to be studied.
In our study, use of a combination of parallel imaging and short interecho
spacing that allowed an echo-train length of 120 and a -90° restore pulse
to shorten spin relaxation time led to significant reduction in imaging time,
allowing near isotropic voxel size approaching 1 mm3 within a
breath-hold. To our knowledge, this technique had not been reported in the
literature at the time of our study. Our study showed that with this
technique, the quality of 3D T2-weighted TSE images is comparable to or better
than that of conventional 2D thick-slab T2-weighted TSE images. A benefit of
obtaining images with nearly isotropic voxels ( 1 mm in all three
dimensions) is the ability to perform 3D reconstructions with a variety of
imaging tools to depict complex biliary and pancreatic relationships (Fig.
6A,
6B,
6C). Anticipated advantages
include improved detection of small calculi and masses and improved
preoperative assessment of biliary anatomic features in potential liver donors
[21].
|
|
|
The free-breathing 3D method produced more artifacts than the breath-hold 3D approach, mainly because of blurring from respiratory motion (Fig. 5A, 5B, 5C). The respiratory artifacts may have been related to an erratic breathing pattern in some patients, which can be possibly improved by coaching or breathing feedback. Although it had fewer artifacts, the breath-hold 3D technique had a trend toward lower relative signal intensity and contrast compared with the free-breathing technique. The cause was a shorter acquisition time, which likely contributed to the trend toward slightly lower scores for delineation of pancreatic ductal anatomic and pathologic features. In addition, in the patient population with limited breath-hold capability, 2D and free-breathing 3D techniques are probably more robust than the breath-hold 3D technique. Continued advancements in MRI technology, including improved coils, higher SNR at 3 T, and a higher parallel acquisition technique factor, may make the breath-hold technique the method of choice.
There were recognized limitations to this study. First, we compared 3D imaging only with an established thick-slab heavily T2-weighted method [12, 22-24] and not conventional 2D single-shot half-Fourier TSE images. Because our aim was to compare image quality of two methods that produce comparable contrast, we believed this comparison would be more meaningful. In addition, we assessed only image quality rather than diagnostic accuracy of the 3D methods because correlative imaging and intraoperative findings were not available for our study population. Whether improved image quality translates into increased accuracy remains to be determined. Although the 3D methods were not differentiated by the blinded reviewers, the reviewers did differentiate 2D from 3D images because the source images and maximum-intensity-projection images were provided for the 3D data sets. Future prospective studies with pathologic correlation may help eliminate this potential for bias. In addition, to assess the intrinsic quality of the breath-hold 3D MRCP sequence, we included only patients who could perform a 30-second breath-hold. This decision may have biased performance of the free-breathing 3D MRCP sequence because these patients tended to breathe slowly and rhythmically (ideal for respiratory triggering). The free-breathing sequence may not perform as well in patients who breathe rapidly and irregularly.
In summary, we showed the feasibility of a 3D heavily T2-weighted MRCP technique of parallel imaging that provides 1-mm voxels and better delineates pancreaticobiliary anatomic features than do conventional heavily T2-weighed 2D techniques. A respiration-triggered version of the 3D sequence requires acquisition times of approximately 2 minutes and provides superior image quality but also more artifacts than a 30-second breath-hold version. We anticipate that future prospective studies will verify that the improvements in image quality with both 3D imaging methods will translate into better diagnostic accuracy and improved patient care.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  K. Hosseinzadeh, A. Furlan, and O. Almusa 2D Thick-Slab MR Cholangiopancreatography: Does Parallel Imaging with Sensitivity Encoding Improve Image Quality and Duct Visualization? Am. J. Roentgenol., June 1, 2008; 190(6): W327 - W334. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Q. Knowlton, A. J. Taylor, M. Reichelderfer, and J. Stang Imaging of Biliary Tract Inflammation: An Update Am. J. Roentgenol., April 1, 2008; 190(4): 984 - 992. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
