AJR 2001; 176:1183-1189
© American Roentgen Ray Society
MR Cholangiography with Volume Rendering
Receiver Operating Characteristic Curve Analysis in Patients with Choledocholithiasis
Hiroshi Kondo1,
Masayuki Kanematsu1,
Yoshimune Shiratori2,
Kyo Itoh3,
Takamichi Murakami4,
Masatoshi Hori4,
Ichiro Yasuda2,
Masayuki Matsuo1,
Hironobu Nakamura4,
Hiroaki Hoshi1 and
Hisataka Moriwaki2
1
Department of Radiology, Gifu University School of Medicine, 40 Tsukasamachi,
Gifu 500-8705, Japan.
2
First Department of Internal Medicine, Gifu University School of Medicine,
Gifu 500-8705, Japan.
3
Department of Radiology, Kyoto University Faculty of Medicine, Kyoto 606-8501,
Japan.
4
Department of Radiology, Osaka University Medical School, Osaka 565-0871,
Japan.
Received September 11, 2000;
accepted after revision October 23, 2000.
Address correspondence to M. Kanematsu.
Abstract
OBJECTIVE. The purpose of our study was to compare observer
performances for the diagnosis of choledocholithiasis using MR cholangiography
with volume-rendered, maximum-intensity-projection, and thick-section
half-Fourier rapid acquisition with relaxation enhancement sequences.
MATERIALS AND METHODS. The images from three types of MR
cholangiography performed on 43 patients with biliary calculi were
retrospectively analyzed. Image review was conducted for two anatomic
compartments (upper biliary tract and common bile duct). A total of 86
compartments, including 19 with bile duct calculi, were reviewed by three
independent off-site gastrointestinal radiologists. Observer performance was
determined by receiver operating characteristic curve analysis. Image quality
was subjectively judged by three radiologists.
RESULTS. Sensitivity was higher with volume-rendered MR
cholangiography (58%) than with thick-section (54%, not significant) and
maximum-intensity-projection MR cholangiography (47%, p < 0.07).
Specificity was higher with volume-rendered MR cholangiography (92%) than with
thick-section (86%, p < 0.03) and maximum-intensity-projection MR
cholangiography (88%, not significant). Accuracy was higher with
volume-rendered MR cholangiography (84%) than with thick-section and
maximum-intensity-projection MR cholangiography (79% for both, not
significant). Observer performance with volume-rendered MR cholangiography
(Az = 0.791-0.952) was better than that with thick-section
(Az = 0.722-0.834) and maximum-intensity-projection MR
cholangiography (Az = 0.771-0.887). Image quality was better with
maximum-intensity-projection MR cholangiography and thick-section MR
cholangiography than with volume-rendered MR cholangiography (p <
0.0001).
CONCLUSION. Observer performance with volume-rendered MR
cholangiography was better than that with maximum-intensity-projection and
thick-section MR cholangiography for the diagnosis of choledocholithiasis.
Volume rendering may be an efficient technique for the reconstruction of MR
cholangiography.
Introduction
MR cholangiopancreatography has developed as an efficient, noninvasive
imaging tool for the diagnosis of pancreaticobiliary diseases. This technique
is feasible in patients in whom sufficient endoscopic retrograde
cholangiopancreatography (ERCP) has failed, in patients who have undergone
gastrectomy or pancreaticoduodenectomy and in whom ERCP cannot be performed,
in patients with acute pancreatitis, and in infants or elderly patients. Until
now, a variety of pulse sequences such as steady-state free precession
[1,
2], two-dimensional fast
spin-echo
[3,4,5,6,7,8,9],
three-dimensional fast spin-echo
[10,
11], and half-Fourier rapid
acquisition with relaxation enhancement (RARE)
[12,13,14]
sequences have been described as optimal imaging sequences for MR
cholangiopancreatography. Recent studies have described the clinical
usefulness of MR cholangiopancreatography performed with half-Fourier RARE
sequences with or without image reconstruction using the
maximum-intensity-projection technique
[15,
16]. MR
cholangiopancreatography reconstructed from sequential coronal source images
3-5 mm thick allows complete visualization of the pancreaticobiliary ducts and
enables a cinematic observation by spinning the images. Reconstructed MR
cholangiopancreatography provides images that can be reviewed in sequence on
the computer monitor, which allows viewing the pancreaticobiliary duct system
rotating in space and improves the observers' ability to recognize the
anatomic architecture and diseased site.
Researchers have described the usefulness of the volume-rendering technique
for reconstructing three-dimensional images of the tracheobronchial tree
[17], colorectal polyps
[18], and splanchnic vessels
[19] with CT images. However,
to our knowledge no previous studies have examined the usefulness of
volume-rendering techniques for MR cholangiography. We assessed the usefulness
of volume-rendered MR cholangiography in patients with choledocholithiasis by
comparing observer performance and image quality of MR cholangiography with
maximum-intensity-projection, volume-rendered, and thick-section half-Fourier
RARE sequences.
Materials and Methods
Patient Study
During the 16-month period between June 1997 and September 1998, 248
consecutive patients who were suspected of having pancreaticobiliary disease
on the basis of findings on previously performed transabdominal sonography,
CT, or serologic tests underwent MR imaging at our department. All patients
understood that the MR examination was primarily for clinical diagnosis and
secondarily for radiologic research. Two hundred five patients without proof
of the presence of biliary calculi after the radiologic workup with unenhanced
helical CT (5- to 7-mm collimation, 120 kVp, 200-220 mAs), transabdominal
sonography (3.75-MHz convex transducer), ERCP, or follow-up imaging were
excluded from the study population because the purpose of our study was to
compare the observer performances among the three types of images. The
remaining 43 patients, including 21 men and 22 women (age range, 39-94 years;
mean age, 65 years), formed the study population. We included patients with
cholecystlithiasis alone because radiologists commonly need to further
scrutinize for the presence of bile duct calculi on MR images when gallbladder
calculi are seen. Of the 43 patients, 25 who were suspected of having
choledocholithiasis underwent ERCP; 17 of these patients were diagnosed as
having choledocholithiasis, and eight as having cholecystlithiasis alone. All
remaining 18 patients with cholecystlithiasis alone were examined with
transabdominal sonography (3.75-MHz convex transducer) and thin-collimation CT
(5-mm collimation, 120 kVp, 200-220 mAs), and were determined not to have
choledocholithiasis.
MR Imaging Techniques
MR imaging was performed using a 1.5-T MR unit (Signa Horizon; General
Electric Medical Systems, Milwaukee, WI) and phased array body multicoil
(Torso-array coil; General Electric Medical Systems). The imaging protocol for
MR cholangiography comprised T2-weighted imaging with chemical shift selective
fat-suppressed respiratory-triggered fast spin-echo imaging (effective TR
range/effective TE range, 7500-15,000/252-255; echo train length, 12-18; 2
signals averaged; matrix size, 256 x 192; kHz, ±16; field of
view, 24 x 24 cm; 20-28 sections per 3.5-5.5 min) and non-fat-suppressed
breath-hold half-Fourier single-shot fast spin-echo imaging (effective
TR/effective TE, infinite/81,034; echo train length, 136; half-Fourier
acquisition; matrix size, 256 x 256; kHz, ±31.3; field of view,
24 x 24 cm; acquisition time, 1 section per second).
Respiratory-triggered fast spin-echo images for the reconstruction of MR
cholangiography were obtained in the coronary plane with 3-mm section
thickness and no intersection gap. Thick-section MR cholangiography images
were obtained in the coronal, oblique, and sagittal planes with a section
thickness of 20-50 mm. Six images at every 30° over 180° rotation with
the z-axis were obtained. Thinner section thickness was used for
anterior images to eliminate signals from gastric juice and spinal fluid, and
thicker section thickness for oblique images to cover areas as broad as
possible.
Reconstruction of MR cholangiography was done with a commercially available
workstation (Advantage Windows version 2.0; General Electric Medical Systems)
for maximum-intensity-projection reconstruction, and another commercially
available workstation (Magic View version 3.1; Siemens, Erlangen, Germany) for
volume-rendering reconstruction. The z-axis was selected for
cinematic three-dimensional observation at every 10° over 180°
rotation, resulting in 19 images each for the maximum-intensity-projection and
volume-rendering techniques. To generate volume-rendered MR cholangiography,
parameter selection was based on our preliminary findings (Kanematsu M, Kondo
H, Matsuo M, unpublished data). The position of the trapezoid was determined
so that the voxels located at the interface between the bile and calculi were
displayed. The trapezoid defined by window width and level was positioned on a
histogram of signal intensity of source images displayed on the computer
monitor (Fig. 1), in which the
x-axis represented the signal-intensity value, which ranged from
-1000 to 3096 H, and the y-axis represented the degree of opacity
expressed as a percentage, at which 0% opacity indicates completely
transparent and 100% opacity indicates completely opaque. We empirically used
four preset points to determine the position of the trapezoid: -980 to -680 H
at 0% opacity level, and -900 to -700 H at 100% opacity level
(Fig. 1). The window level was
always chosen at the midpoint of the window width at the 100% opacity level.
For the remaining volume-rendering parameters, the parenchymal opacification
was set at 15% and an unshaded algorithm was used. The 19 images each of the
maximum-intensity-projection and volume-rendered MR cholangiograms were
formatted into two pieces of 4 x 5 formatted film using a laser imager.
It typically required 5 min to produce the maximum-intensity-projection or
volume-rendered MR cholangiogram for one patient.

View larger version (22K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 1. Histogram shows opacity percentage plotted against Hounsfield
value. Trapezoid was placed so that voxels located at interface between bile
and calculi were seen. Four fixed points were used to determine position of
trapezoid: -980 to -680 H at 0% opacity level, and -900 to -700 H at 100%
opacity level.
|
|
Image Analysis
Three off-site radiologists, who had served mainly as gastrointestinal
radiologists for 7-17 years (mean, 12.5 years) with experience in interpreting
MR images of the pancreaticobiliary system were invited from other
institutions to conduct the image review. These radiologists independently
reviewed maximum-intensity-projection, volume-rendered, and thick-section MR
cholangiograms obtained in the 43 patients. They knew that the patients were
referred for assessment of suspected biliary calculi but did not have any
other information about the patients' histories.
The image review was conducted for two anatomic compartments: upper biliary
tracts (intrahepatic bile ducts, right and left hepatic ducts, and common
hepatic duct) and the common bile duct. We performed the image review on a
compartment-by-compartment basis because correct localization of biliary
calculi is necessary when determining treatment options such as endoscopic
lithotripsy, laparoscopic cholecystectomy, laparocystectomy, or
laparolithotripsy, and because our objective was to compare observer
performance with the three imaging sequences by means of receiver operating
characteristic curve analysis.
The three radiologists independently reviewed 86 anatomic compartments in
43 patients, including 19 compartments that harbored bile duct calculi. Images
were reviewed in alphabetic order according to the patients' names, but the
order in which images obtained with the three imaging sequences were reviewed
was randomized. In other words, images from all patients were reviewed at a
single session, but only the images obtained with one of the three imaging
sequences were reviewed for a given patient at that session. The other types
of images were reviewed at subsequent sessions. To minimize learning bias, the
name, age, identification number, and imaging parameters were masked.
The radiologists recorded the size and site of the visible area of the
signal-intensity defect that was considered to be caused by bile duct calculi,
and indicated whether the presence of bile duct calculi could be ascertained
for each anatomic compartment. Each radiologist allocated one of five
confidence levels (1 = definitely absent, 2 = probably absent, 3 = equivocal,
4 = probably present, 5 = definitely present) to each compartment. When a
large biliary calculus was located over two compartments, the radiologist was
asked to consider only the compartment in which most of the calculus was
located and to assess the probability of other bile duct calculi in another
compartment. The radiologists were instructed to indicate a score of 1 when no
signal-intensity decrease or defect was seen; a score of 3 when the
signal-intensity decrease or defect was subtle, ill-defined, or not
gravity-dependent; and a score of 5 when the signal-intensity decrease or
defect was discrete, well-circumscribed, or gravity-dependent. Scores of 2 and
4 were allocated according to the radiologist's subjective judgment.
Futhermore, each radiologist evaluated image quality in terms of image
sharpness, image contrast, depiction of fine structures, and recognition of
anatomy using a five-point scale (1 = poor, 2 = fair, 3 = good, 4 = very good,
5 = excellent). A "poor" score was assigned when the image could
not be interpreted because of image degradation, a "good" score
was assigned when the image degradation was present but did not markedly
preclude interpretation, and an "excellent" score was assigned
when the image was virtually free from image degradation. Scores of
"fair" and "very good" were allocated according to the
radiologist's subjective judgment.
Statistical Analysis
Sensitivity for detection of bile duct calculi was determined using the
number of anatomic compartments assigned a score of 4 or 5 (i.e., probably
present or definitely present) of the total number of 19 compartments with
bile duct calculi. Specificity was determined using the number of compartments
assigned a score of 1 or 2 (i.e., definitely absent or probably absent) of the
total number of 67 compartments without bile duct calculi. Sensitivity and
specificity were compared using the McNemar test. Accuracy was compared with
the chi-square test.
For each imaging sequence, a receiver operating characteristic curve was
fitted to each observer's confidence rating using a maximum-likelihood
estimation with the ROCKIT 0.9.1B program (Metz, University of Chicago,
Chicago, IL) [20]. Observer
performance was estimated by calculating the area under the receiver operating
characteristic curve (Az). Differences between the areas under the
receiver operating characteristic curves were tested using the area test with
univariate z-score test.
To assess interobserver variability in interpreting images, kappa
statistics for multiple observers were used to measure the degree of agreement
among the three observers. We used the nonweighted kappa statistics with
binary data defined in terms of the presence (i.e., definitely present,
probably present, or equivocal) or absence (i.e., probably absent or
definitely absent) of calculi in a compartment. Kappa values of up to 0.40
indicated positive but poor agreement, values of 0.41-0.75 indicated good
agreement, and values greater than 0.75 indicated excellent agreement.
Results
In the 43 patients, biliary calculi were located in the gallbladder alone
in 26 patients, upper biliary tract alone in one, common bile duct alone in
five, gallbladder and upper biliary tract in one, gallbladder and common bile
duct in eight, upper biliary tract and common bile duct in one, and all
compartments in one.
The sensitivity, specificity, and accuracy for detection of bile duct
calculi are listed in Table 1.
Sensitivity with volume-rendered MR cholangiography was marginally higher than
that with maximum-intensity-projection MR cholangiography in the total data
(p < 0.07) (Figs.
2A,2B,2C,3A,3B,3C,4A,4B,4C).
Specificity with volume-rendered MR cholangiography was marginally higher than
that with thick-section (p < 0.06) and
maximum-intensity-projection (p < 0.07) MR cholangiography for
radiologist 2, and was significantly higher than that with thick-section MR
cholangiography in the total data (p < 0.03). Accuracy with
volume-rendered MR cholangiography was significantly higher than that with
maximum-intensity-projection MR cholangiography (p < 0.05) and
marginally higher than that with thick-section MR cholangiography (p
< 0.07) for radiologist 2, and was marginally higher than that with
maximum-intensity-projection and thick-section MR cholangiography in the total
data (p < 0.09).

View larger version (121K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2A. 58-year-old man with common bile duct calculus
(arrows, A-C). Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that visibility of calculus is comparable for all
three types of images. However thick-section MR cholangiography is less
blurred as result of 1-sec data acquisition during respiratory suspension.
|
|

View larger version (112K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2B. 58-year-old man with common bile duct calculus
(arrows, A-C). Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that visibility of calculus is comparable for all
three types of images. However thick-section MR cholangiography is less
blurred as result of 1-sec data acquisition during respiratory suspension.
|
|

View larger version (78K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 2C. 58-year-old man with common bile duct calculus
(arrows, A-C). Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that visibility of calculus is comparable for all
three types of images. However thick-section MR cholangiography is less
blurred as result of 1-sec data acquisition during respiratory suspension.
|
|

View larger version (150K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3A. 48-year-old woman with gallbladder and common bile duct
calculi (arrows, B). Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that image in B shows both calculi more
clearly than other types of images.
|
|

View larger version (129K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3B. 48-year-old woman with gallbladder and common bile duct
calculi (arrows, B). Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that image in B shows both calculi more
clearly than other types of images.
|
|

View larger version (137K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 3C. 48-year-old woman with gallbladder and common bile duct
calculi (arrows, B). Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that image in B shows both calculi more
clearly than other types of images.
|
|

View larger version (135K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 4A. 68-year-old man with numerous calculi in upper biliary tract,
common bile duct, and gallbladder. Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered, (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that calculi in upper biliary tract (straight
arrows, B and C) are better shown in B and C
than in A. Note that image in B clearly shows calculus in upper
biliary tract that is superimposed by signal intensity of gastric juice
(curved arrow, B).
|
|

View larger version (139K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 4B. 68-year-old man with numerous calculi in upper biliary tract,
common bile duct, and gallbladder. Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered, (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that calculi in upper biliary tract (straight
arrows, B and C) are better shown in B and C
than in A. Note that image in B clearly shows calculus in upper
biliary tract that is superimposed by signal intensity of gastric juice
(curved arrow, B).
|
|

View larger version (125K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 4C. 68-year-old man with numerous calculi in upper biliary tract,
common bile duct, and gallbladder. Anterior MR cholangiograms with
maximum-intensity-projection (A), volume-rendered, (B), and
thick-section half-Fourier rapid acquisition with relaxation enhancement
(C) sequences show that calculi in upper biliary tract (straight
arrows, B and C) are better shown in B and C
than in A. Note that image in B clearly shows calculus in upper
biliary tract that is superimposed by signal intensity of gastric juice
(curved arrow, B).
|
|
The receiver operating characteristic curves for each radiologist are shown
in Figures
5,6,7.
Observer performance for detection of bile duct calculi was marginally greater
with volume-rendered MR cholangiography than with maximum-intensity-projection
MR cholangiography for radiologists 2 and 3 (p < 0.08 for both).
Observer performance was significantly greater with volume-rendered MR
cholangiography than with thick-section MR cholangiography for radiologist 2
(p < 0.04). No significant difference in observer performance was
found between maximum-intensity-projection and thick-section MR
cholangiography. Kappa values were 0.44, 0.60, and 0.58 for
maximum-intensity-projection, volume-rendered, and thick-section MR
cholangiography, respectively. Good agreement was obtained among the
radiologists.

View larger version (16K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 5. Graph shows receiver operating characteristic curves for
detection of bile duct calculi by radiologist 1 for
maximum-intensity-projection ([UNK]) (Az = 0.771),
volume-rendered ([UNK]) (Az = 0.791), and thick-section
( ) (Az = 0.722) MR cholangiograms. Observer
performance with maximum-intensity-projection and volume-rendered MR
cholangiography tends to exceed that with thick-section MR
cholangiography.
|
|

View larger version (18K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 6. Graph shows receiver operating characteristic curves for
detection of bile duct calculi by radiologist 2 for
maximum-intensity-projection ([UNK]) (Az = 0.887),
volume-rendered ([UNK]) (Az = 0.952), and thick-section
( ) (Az = 0.834) MR cholangiograms. Observer
performance with volume-rendered MR cholangiography significantly (p
< 0.04) exceeds that with thick-section MR cholangiography, and marginally
(p < 0.08) exceeds that with maximum-intensity-projection MR
cholangiography.
|
|

View larger version (17K):
[in this window]
[in a new window]
[as a PowerPoint slide]
|
Fig. 7. Graph shows receiver operating characteristic curves for
detection of bile duct calculi by radiologist 3 for
maximum-intensity-projection ([UNK]) (Az = 0.777),
volume-rendered ([UNK]) (Az = 0.848), and thick-section
( ) (Az = 0.830) MR cholangiography. Observer
performance with volume-rendered MR cholangiography marginally (p
< 0.08) exceeds that with maximum-intensity-projection MR
cholangiography.
|
|
The mean scores for the image quality are shown in
Table 2. Image quality was
significantly better with maximum-intensity-projection MR cholangiography and
thick-section MR cholangiography than with volume-rendered MR cholangiography
(p < 0.0001) (Figs.
2A,2B,2C,3A,3B,3C,4A,4B,4C).
Discussion
The maximum-intensity-projection technique is one of the primary means of
displaying three-dimensional images and has been used for reconstructing MR
cholangiopancreatography. However, the maximum-intensity-projection technique
uses only approximately 10% of data
[21], and the limitations of
the maximum-intensity-projection technique are that overlapping ducts are not
depicted as areas of increased density, as occurs with conventional
cholangiography, and no information on depth is available. These limitations
make evaluation of a single maximum-intensity-projection image difficult.
Furthermore, small polyps or calculi are masked if brighter voxels in source
images without disease are projected instead of the darker voxels caused by
disease. The volume-rendering technique is a novel reconstruction algorithm
that is frequently used in clinical practice as a result of recent advances in
computer graphics software and hardware. With volume rendering, which uses
information from all available voxels not just those at the surface of the
objects, some voxels, are rendered opaque whereas others are transparent.
Furthermore, the reconstruction algorithm and display technology used in
volume rendering yield a number of advantages over
maximum-intensity-projection or shaded-surfacedisplay techniques
[22].
In our study, analysis of sensitivity showed a trend toward the superiority
of volume-rendered MR cholangiography over maximum-intensity-projection MR
cholangiography, although these two types of MR cholangiography were
reconstructed from the identical source images. With the
maximum-intensity-projection technique, low-signal-intensity voxels caused by
calculi were often masked by brighter voxels caused by bile around the
lesions. Meanwhile, with the volume-rendering technique, all voxels of source
images were transparent to some degree, and the light reflection from voxels
inside the structures could be observed, resulting in its superior ability to
reveal bile duct calculi (Figs.
3A,3B,3C
and
4A,4B,4C).
Specificity with volume-rendered MR cholangiography was significantly
superior to that with thick-section MR cholangiography. Because thick-section
MR cholangiography takes advantage of the partial volume-averaging effect, the
thicker the imaging section, the less the image contrast, possibly decreasing
the conspicuousness of small calculi. Because the radiologists knew of the low
image contrast with thick-section images, they might have recognized the small
areas of subtly decreased signal intensity caused by artifacts as areas of
disease.
The receiver operating characteristic curve analysis indicated a trend
toward the superiority of volume-rendered MR cholangiography over
maximum-intensity-projection and thick-section MR cholangiography. The
capability of volume-rendered MR cholangiography to show inside structures
presumably yielded the higher observer performance. Observer performance with
volume-rendered MR cholangiography was significantly superior to that with
thick-section MR cholangiography even though volume-rendered MR
cholangiography was interpreted without referring to its source images in this
study. Had the volume-rendered MR cholangiography and its source images been
reviewed together, the observer performance might have been better than that
yielded by volume-rendered MR cholangiography alone.
Image quality was generally inferior with volume-rendered MR
cholangiography. Image sharpness deteriorated with volume-rendered and
maximum-intensity-projection MR cholangiography, especially in patients whose
irregular respiration caused blurring of the source images. To remedy this,
the source images may be obtained during breath-hold using a half-Fourier RARE
sequence. However, the respiratory-triggered fast spin-echo sequence was
preferred because an adequate number of thin sections covering the entire
biliary system could be obtained, and the contrast was well preserved even
with 3-mm section thickness. We frequently experienced substantial blurring of
source images obtained with a breath-hold half-Fourier RARE sequence, which
may have been caused by the patient's imperfect respiratory suspension. Image
contrast was inferior with volume-rendered MR cholangiography, probably
because we set the opacity at 15%; 85% of the light penetrated the voxels,
resulting in transparent images that were often ghostlike compared with
maximum-intensity-projection MR cholangiograms. Regarding the depiction of
fine structures, the peripheral intrahepatic bile ducts were poorly depicted
with volume-rendered MR cholangiography compared with
maximum-intensity-projection MR cholangiography. The low setting of opacity
percentage might be related to the poor depiction of fine anatomic structures
with volume-rendered MR cholangiography.
There are some limitations to this study. The 17 patients with
choledocholithiasis and eight with cholecystlithiasis alone underwent ERCP for
the confirmation of disease, but the remaining 18 patients with
cholecystlithiasis alone did not. Ideally, proof of cholelithiasis is obtained
at ERCP or intraoperative direct cholangiography. However, noninvasive MR
cholangiography is often substituted for diagnostic ERCP, and intraoperative
cholangiograms are not obtained in the event that laparoscopic cholecystectomy
is performed. Thirty-six of the 43 patients in the current study had
cholecystlithiasis; this high prevalence of cholecystlithiasis might have led
the radiologists, who were unaware of the patients' histories, to an
interpretation bias toward the frequent positive findings of
choledocholithiasis. The significance in sensitivity and specificity in our
study was limited because of the exclusion of all patients without biliary
calculi.
The three radiologists reviewed reconstructed MR cholangiograms alone,
without referring to their source images. Although it is common to refer to
source images of MR cholangiography or transaxial T2-weighted images in the
clinical setting, we deliberately avoided having the radiologists review the
source images to achieve a net comparison of volume-rendered and
maximum-intensity-projection MR cholangiography. We could not completely blind
the off-site radiologists to the types of imaging sequences because each
sequence had specific imaging features. The off-site radiologists invited from
other institutions did not have any clinical experience with volume-rendered
MR cholangiography, whereas they had routinely interpreted
maximum-intensity-projection and thick-section MR cholangiography at their
institutions. Such unfamiliarity with volume-rendered MR cholangiography might
have affected the observer performance and assessment of image quality.
However, the diagnostic accuracy with volume-rendered images might have been
higher had the radiologists been accustomed to interpreting volume-rendered
images. Finally, although we limited our study to the evaluation of MR
cholangiography for the detection of choledocholithiasis, the usefulness of
volume-rendered MR cholangiopancreatography for the diagnosis of
pancreaticobiliary disease caused by malignancy, inflammation, or congenital
disorder should be further evaluated.
In conclusion, observer performance for volume-rendered MR cholangiography
tended to exceed that of maximum-intensity-projection MR cholangiography in
sensitivity and exceeded thick-section MR cholangiography in specificity in
the diagnosis of choledocholithiasis. The receiver operating characteristic
study showed a trend toward the superiority of volume-rendered MR
cholangiography over maximum-intensity-projection and thick-section MR
cholangiography. The volume-rendering technique is recommended rather than the
maximum-intensity-projection technique for reconstruction of MR
cholangiography for the diagnosis of choledocholithiasis.
Acknowledgments
We thank Hirokazu Osakabe of Siemens-Asahi Medical Technologies, Tokyo,
Japan, for technical advice.
References
-
Wallner BK, Schumacher KA, Weidenmaier W, Friedrich JM. Dilated
biliary tract: evaluation with MR cholangiography with a T2-weighted CE-FAST
sequence. Radiology
1991;181:805
-808[Abstract/Free Full Text]
-
Morimoto K, Shimoi M, Shirakawa T, et al. Biliary obstruction:
evaluation with three-dimensional MR cholangiography.
Radiology
1992;183:578
-580[Abstract/Free Full Text]
-
Nall-Craggs MA, Allen CM, Owens CM, et al. MR cholangiography:
clinical evaluation in 40 cases. Radiology
1993;189:423
-427[Abstract/Free Full Text]
-
Ishizaki Y, Wakayama T, Okada Y, Kobayashi T. Magnetic resonance
cholangiography for evaluation of obstructive jaundice. Am J
Gastroenterol 1993;88:2072
-2077[Medline]
-
Guilbaud L, Bret PM, Reinhold C, Atri M, Barkun ANG. Diagnosis of
choledocholithiasis: value of MR cholangiography. AJR
1994;163:847
-850[Abstract/Free Full Text]
-
Macaulay SE, Schulte SJ, Sekijima JH, et al. Evaluation of
non-breath-hold MR cholangiography technique.
Radiology
1995;196:227
-232[Abstract/Free Full Text]
-
Guilbaud L, Bret PM, Reinhold C, Atri M, Barkun AN. Bile duct
obstruction and choledocholithiasis: diagnosis with MR cholangiography.
Radiology
1995;197:109
-115[Abstract/Free Full Text]
-
Chan Y, Chan ACW, Lam WWM, et al. Choledocholithiasis: comparison
of MR cholangiography and endoscopic retrograde cholangiography.
Radiology
1996;200:85
-89[Abstract/Free Full Text]
-
Takehara Y, Ichijo K, Tooyama N, et al. Breathhold MR
cholangiopancreatography with a long-echo-train fast spin-echo sequence and a
surface coil in chronic pancreatitis. Radiology
1994;192:73
-78[Abstract/Free Full Text]
-
Soto JA, Barish MA, Yucel EK, et al. Pancreatic duct: MR
cholangiopancreatography with a three-dimensional fast spin-echo technique.
Radiology
1995;196:459
-464[Abstract/Free Full Text]
-
Barish MA, Yucel EK, Soto JA, Chuttani R, Ferrucci JT. MR
cholangiopancreatography: efficacy of three-dimensional turbo spin-echo
technique. AJR
1995;165:295
-300[Abstract/Free Full Text]
-
Regan F, Smith D, Khazan R, et al. MR cholangiography in biliary
obstruction using half-Fourier acquisition. J Comput Assist
Tomogr 1996;20:627
-632[Medline]
-
Miyazaki T, Yamashita Y, Tsuchigame T, Yamamoto H, Urata J,
Takahashi M. MR cholangiopancreatography using HASTE (half-Fourier acquisition
single-shot turbo spin-echo) sequences. AJR
1996;166:1297
-1303[Abstract/Free Full Text]
-
Ichikawa T, Nitatori T, Hachiya J, Mizutani Y. Breath-held MR
cholangiopancreatography with half-averaged single shot hybrid rare
acquisition with relaxation enhancement sequence: comparison of fast GRE and
SE sequences. J Comput Assist Tomogr
1996;20:798
-802[Medline]
-
Ernst O, Calvo M, Sergent G, Mizrahi D, Carpentier F. Breath-hold
MR cholangiopancreatography using a HASTE sequence: comparison of single-shot
and multislice acquisition techniques. AJR
1997;169:1304
-1306[Free Full Text]
-
Irie H, Honda H, Tajima T, et al. Optimal MR
cholangiopancreatographic sequence and its clinical application.
Radiology
1998;206:379
-387[Abstract/Free Full Text]
-
Remy-Jardin M, Remy J, Artaud D, Fribourg M, Duhamel A. Volume
rendering of the tracheobronchial tree: clinical evaluation of bronchographic
images. Radiology
1998;208:761
-770[Abstract/Free Full Text]
-
McFarland EG, Brink JA, Loh J, et al. Visualization of colorectal
polyps with spiral CT colography: evaluation of processing parameters with
perspective volume rendering. Radiology
1997;205:701
-707[Abstract/Free Full Text]
-
Johnson PT, Health DG, Kuszyk BS, Fishman EK. CT angiography with
volume rendering: advantages and applications in splanchnic vascular imaging.
Radiology
1996;200:564
-568[Abstract/Free Full Text]
-
Metz CE. ROC methodology in radiologic imaging. Invest
Radiol 1986;21:720
-723[Medline]
-
Heath DR, Soyer P, Kuszyk B, et al. Three-dimensional spiral CT
during arterial portography: comparison of 3-D rendering techniques.
Radio-Graphics
1995;15:1001
-1011[Abstract]
-
Rubin GD, Dake MD, Semba CP. Current status of three-dimensional
spiral CT scanning for imaging the vasculature. Radiol Clin North
Am 1995;33:51
-70[Medline]

CiteULike
Complore
Connotea
Del.icio.us
Digg
Reddit
Technorati What's this?