AJR 2005; 185:700-703
© American Roentgen Ray Society
Suspected Pancreatic Cancer: Evaluation by Dynamic Gadolinium-Enhanced 3D Gradient-Echo MRI
Katherine R. Birchard1,
Richard C. Semelka1,
W. Brian Hyslop1,
Alfonso Brown2,
Diane Armao1,
Zeynep Firat1 and
Georgeta Vaidean3
1 Department of Radiology, University of North Carolina Hospitals, 101 Manning
Dr., 2006 Old Clinic Bldg., Chapel Hill, NC 27599-7510.
2 Department of Medicine, University of North Carolina Hospitals, Chapel Hill,
NC 27599-7510.
3 School of Public Health, University of North Carolina at Chapel Hill, Chapel
Hill, NC 27599-7510.
Received July 15, 2004;
accepted after revision October 6, 2004.
Address correspondence to R. C. Semelka.
Abstract
OBJECTIVE. The objective of our study was to determine the
sensitivity and specificity of gadolinium-enhanced 3D gradient-echo (GRE) MR
images in the detection of pancreatic cancer.
CONCLUSION. Dynamic gadolinium-enhanced 3D GRE images are both
sensitive and specific in the detection of pancreatic cancer. Our study shows
that the identification of pancreatic cancer using these images can be
performed with a high degree of confidence and accuracy, making them very
useful in the detection of pancreatic cancer.
Introduction
Most patients with pancreatic cancer have advanced disease at the
time of diagnosis [1]. Early
detection of pancreatic cancer may allow the chance for complete surgical
resection. In the past, MRI of the pancreas has been limited by image
artifacts caused by aortic pulsation, breathing motion, and bowel peristalsis.
Breath-hold spoiled gradient-echo (SGE) sequences, along with fat suppression
and dynamic gadolinium-enhanced images, have reduced artifacts and resulted in
images of sufficient quality to image the pancreas
[2,
3]. Recently, the development
of a 3D gradient-echo (GRE) sequence, volumetric interpolated breath-hold
examination (VIBE, Siemens Medical Solutions), has led to improved
visualization of the pancreas. This sequence allows high spatial resolution
and high contrast resolution with decreased phase artifact and decreased
partial volume effects [4]. 3D
GRE, therefore, may be an excellent tool for evaluating the pancreas and
detecting pancreatic cancer. To evaluate its usefulness, we retrospectively
reviewed all MRI studies that included these images of patients referred for
examination because of clinically suspected pancreatic cancer.
Materials and Methods
Patients
The study reviewed MRI examinations with gadolinium-enhanced 3D GRE images
of 57 patients who, between June 2002 and August 2003, were referred for
examination because of clinical suspicion of pancreatic cancer. The use of MRI
was at the discretion of the clinician. Before MRI examination, all patients
had clinical symptomatology that included, but was not limited to, weight
loss, abdominal pain, jaundice, fatigue, and early satiety. No patient had
more than one MRI examination included in the study. Seventeen patients
underwent CT before MRI examination. The mean patient age was 58 years (range
= 32-84 years). All patients were followed for a minimum of one year post-MRI
examination (mean = 17 months, range = 12-26 months). All MRI studies were
performed for clinical indications, so institutional review board (IRB)
approval was obtained for case review only. Signature waiver was also obtained
in accordance with IRB regulations and the Health Insurance Portability and
Privacy Act (HIPPA).
MRI Technique
MRI examinations of the abdomen were performed on a 1.5 T system (Sonata,
Siemens Medical Solutions) using a set protocol including unenhanced
T1-weighted images acquired as breath-hold SGE (TR = 120-170 msec, TE =
4.0-4.5 msec, flip angle 80-90°) and T2-weighted HASTE sequence (TR =
infinite, effective TE = 90 msec, 2-3 acquisitions). Section thickness was 5-7
mm and matrix size was 128-192 x 256 (phase frequency encoding) for all
sequences. Gadolinium was administered using a power injector (Spectris MR
injector, Medrad) bolus of 0.1 mmol/kg of gadolinium chelate (Omniscan,
Nycomed) at 2 mL/sec in all patients. T1-weighted 3D GRE VIBE images were
acquired in the axial plane unenhanced and immediately, 45, and 90 sec after
the administration of contrast material, using TR/TE, 4.78/2.27; flip angle,
10°; thickness, 3 mm; fat suppression; and field of view, 300 mm. All
studies were obtained during breath-hold with an acquisition time ranging from
15-22 sec, due to varying field of view, matrix, and number of partitions that
were individually modified according to patient stature. No bowel preparation
was used in any study.
Image Interpretation
All MRI examinations were retrospectively independently interpreted by two
experienced radiologists with fellowship training in body MRI. All MRI
examinations were of diagnostic quality. Patient clinical history and the
original MRI interpretation were withheld from the reviewing radiologists.
Pancreatic neoplasm size, location, associated findings, and evidence of
nonneoplastic pancreatic disease were recorded where applicable. Reviewers
based their diagnoses of pancreatic carcinoma on detection of a mass with
definable margins that enhanced less than background pancreatic tissue on
immediate postgadolinium images and also exhibited interruption of the normal
marbled texture of pancreatic parenchyma. For each MRI examination, reviewers
rated their confidence as to the presence of pancreatic cancer on the
gadolinium-enhanced 3D GRE images using a 5-point scale where 1 represents
very low confidence for pancreatic cancer (no cancer present), 2 represents
low confidence for pancreatic cancer (unlikely that cancer is present), 3
represents intermediate confidence for pancreatic cancer (equivocal), 4
represents high confidence for pancreatic cancer (likely that cancer is
present), and 5 represents very high confidence for pancreatic cancer (cancer
present). Reviewers recorded which sequence best visualized the neoplasm.
Correlation with subsequent clinical records, surgical reports, and pathology
reports was then made to determine the correctness of all MRI
interpretations.
Statistical Analysis
Our study was designed and analyzed as recommended by the Standards for
Reporting Diagnostic Accuracy Steering Group (STARD)
[5]. Weighted kappa statistics
were used to measure the degree of approximate agreement between the two
radiologists (interobserver reproducibility) using SAS software (release 8.0,
SAS Institute). Sensitivity, specificity, likelihood ratios, and 95%
confidence intervals were calculated using the Evidence-Based Medicine (EBM)
calculator on the University of Toronto website (Center for Evidence-Based
Medicine, Mount Sinai Hospital,
www.cebm.utoronto.ca,
accessed February 2004).
Results
Both reviewers identified a pancreatic neoplasm in 27 of 57 patients with
high or very high confidence for pancreatic cancer
(Fig. 1 and
Table 1). The mean size of
identified masses was 2.5 cm (range, 1.2-3.1 cm) with eight of the masses
measuring less than 2.0 cm (Fig.
2). Twelve patients had enlarged (> 1 cm) peri-pancreatic lymph
nodes and/or enlarged porta hepatis lymph nodes. Four patients had evidence of
liver metastases. The reviewers' measurements did not differ by more than 25%
on any MRI examination. Of the 27 patients, 18 were found to have a pancreatic
adenocarcinoma by fine needle aspiration (n = 4), by endoscopic
biopsy (n = 10), or by open surgical biopsy (n = 4). Three
patients were presumed to have pancreatic cancer based on clinical
symptomatology or increasing tumor size on subsequent imaging. Three patients
had negative (benign) pancreatic histologic studies, and each had unremarkable
clinical follow-up. One patient did not have a biopsy to our knowledge, one
patient died of causes unknown to the authors, and the last patient did not
have adequate follow-up records.
Twenty-eight MRI examinations were interpreted as normal and with a
confidence rating of 1 by both reviewers
(Table 1). Clinical records
showed that 27 of the 28 patients had unremarkable clinical follow-up while
one patient died secondary to hepatic failure. No patient with a study
interpreted as normal was subsequently found, by histology or clinical
follow-up, to have a pancreatic malignancy.
Both reviewers identified other pancreatic diseases (neuroendocrine tumor
and pseudocyst) in two patients with a confidence rating of 1 for pancreatic
cancer (Table 1). Clinical
records correlated with the patients' diseases in both cases. No patient in
this group was subsequently found by histology or clinical follow-up to have a
pancreatic carcinoma.
The pancreatic neoplasms identified with high or very high confidence for
carcinoma shared several imaging characteristics. The neoplasms were
hypovascular with definable margins on unenhanced T1-weighted fat-suppressed
images. There was effacement of the normal lobular architecture of surrounding
pancreatic parenchyma. Abnormal decreased enhancement of the suspicious region
with respect to surrounding parenchyma on dynamic postgadolinium images was
ubiquitous. In 20 of 27 patients, neoplasms were best visualized on
T1-weighted immediate postgadolinium 3D GRE images. In seven of 27 patients,
neoplasms were visualized equally well on both unenhanced T1-weighted
fat-suppressed images and on T1-weighted immediate postgadolinium 3D GRE
images. T2-weighted images did not show neoplasms well in any case because of
lack of intrinsic contrast. Overall, pancreatic neoplasms were best visualized
on immediate postgadolinium images.
Sensitivity of the 3D GRE images was calculated for two cutoff points. If
only examinations with a rating of 5 (very high confidence of pancreatic
carcinoma) were considered "positive," the sensitivity was 85.70%
(95% confidence interval [CI]: 65.40-95.00%) and specificity was 88.90% (95%
CI: 74.70-95.60%). If MRI examinations with ratings of 4 (high) and 5 (very
high) were both considered "positive," sensitivity was 97.70% (95%
CI: 81.50-99.90%) and specificity was 85.10% (95% CI: 70.40-93.20%). The
weighted kappa coefficient was 0.91 (95% CI: 0.85-0.97). Interval likelihood
ratios were also computed (Table
2).
Discussion
Despite advances in chemotherapeutic treatments and surgical techniques,
survival rates for pancreatic cancer remain dismal. The presence of advanced
disease at the time of diagnosis and limitations in early detection contribute
to a high mortality rate [6].
Imaging findings often determine the course of treatment. Dynamic enhanced CT
is well established in the evaluation of pancreatic carcinoma
[7,
8]. However, small, isointense,
non-contour-deforming pancreatic carcinomas are often difficult to detect with
dynamic enhanced CT because of limited soft-tissue resolution
[9]. MRI examinations using
T1-weighted fat-suppressed images have been useful in detecting pancreatic
carcinomas [10]. Both 2D GRE
and dynamic gadolinium-enhanced MR images have been particularly effective in
identifying smaller intrapancreatic tumors, peri-pancreatic extension of
tumor, and vascular invasion
[11]. More recently, 3D GRE
images have been comparable to or better than 2D-GRE in the evaluation of the
liver [12]. Advantages of 3D
GRE images over 2D-GRE images include thinner sections, the ability to
reformat 3D images with near isotropic resolution, lack of degradation by
aortic mirror-type phase artifacts that can obscure the midbody of the
pancreas, and overall diminished motion-induced phase artifact. In one study
comparing gadolinium enhanced 2D and 3D GRE images of the abdomen, 3D GRE
images showed better definition of pancreatic margins
[3].
MRI and CT both perform well in pancreatic lesion detection, with MRI being
slightly better in the assessment of resectability
[13]. Probably because of
their hyperintense enhancement, both dynamic contrast-enhanced CT and MRI have
detected neuroendocrine tumors less than 1 cm in diameter
[14,
15]. The average size of
CT-detected resectable pancreatic carcinomas was 2.6 cm in one study and 3.1
cm in another [16,
17]. In our study, five of the
histologically proven pancreatic adenocarcinomas were less than 2.0 cm in
diameter. It would be of interest to determine the lower limit of detectable
pancreatic carcinoma size on both CT and MRI with surgical and pathologic
correlation.
Three of the MRI examinations were rated as high or very high confidence
for pancreatic cancer but had subsequent nonmalignant histologic studies of
the pancreas. In the first patient, histologic study of endoscopically
obtained biopsy showed only inflammatory change and a diagnosis of focal
pancreatitis was made. Pancreatitis and pancreatic carcinoma both show
abnormal enhancement on contrast-enhanced MR images, but distinguishing
characteristics are at times not discernable due to variance in degree and
timing of enhancement [18]. We
believe that this was the case in our study. The second patient had dilated
pancreatic and biliary ducts on the MR image with fullness in the pancreatic
head. Both reviewers suspected an obstructing ampullary mass. Histologic study
of endoscopically obtained biopsy revealed no evidence of malignancy. The
third patient had a mass in the distal body of the pancreas. Histologic study
of surgical specimen revealed a neuroendocrine tumor.
Statistical analysis showed a high sensitivity and specificity for
detection of pancreatic cancer on 3D GRE images. The intraobserver agreement
was excellent as evidenced by the weighted kappa coefficient (0.91). To
capture the magnitude of the test results, interval likelihood ratios were
also computed (Table 2). Our
data suggests that MRI examinations rated as 5 are 6.17 times more common in
patients with histologically or clinically confirmed pancreatic cancer than in
patients without such confirmation. Those patients with MRI examinations rated
as 4 are 5.14 times more common in patients with histologically or clinically
confirmed pancreatic cancer than in patients without such confirmation.
Our study evaluated patients referred for MRI examination by the clinician.
At our institution, MRI has an established role in the investigation of
pancreatic disease, and direct referral to MRI by clinicians is common. Our
study, similar in design to other recent studies, used clinical, surgical, and
pathologic findings to evaluate the usefulness of the 3D GRE images rather
than directly comparing between two techniques
[19]. A future prospective
study comparing 3D GRE images with either 2D GRE images or MDCT would be
helpful in expanding our findings. Comparison between 2D and 3D GRE images
would, however, require two separate MRI examinations because the critical
sequence is obtained immediately post-gadolinium injection. Limitations of our
study include the retrospective nature of data analysis and use of a specific
patient population. We attempted to control observer bias by blinding
reviewers to patient history and clinical information. Although HASTE images
are only mildly T2-weighted, the sequence is useful for evaluating obstruction
in the common bile and pancreatic ducts. In addition, it would be of interest
to study a random patient population in a prospective manner to determine
usefulness, possibly for screening purposes, of 3D GRE images.
In summary, by using 3D GRE MR images, both reviewers were able to identify
pancreatic cancer with a high degree of confidence and accuracy. In addition,
both reviewers correctly identified normal, malignant, and nonmalignant
pancreatic disease with a high degree of intraobserver agreement. Our data
supports the usefulness of dynamic gadolinium-enhanced 3D GRE MR images in the
detection of pancreatic cancer. The statistical values support its clinical
utility. As more experience is gained with newer MRI sequences and as image
quality continues to improve, MRI is expected to play an increasing role in
the early detection of pancreatic cancer.
References
- Moossa AR, Pancreatic cancer, Ontario, BC:
Decker Inc., 2001: 67-85
- Gohde SC, Toth J, Krestin GP, Debatin JF. Dynamic contrast-enhanced
FMPSPGR of the pancreas: impact on diagnostic performance.
AJR 1997;168:689
-696[Abstract/Free Full Text]
- Obuz F, Dicle O, Coker A, Sagol O, Karademir S. Pancreatic
adenocarcinoma: detection and staging with dynamic MR imaging. Eur
J Radiol 2001;38:146
-150[CrossRef][Medline]
- Rofsky NM, Lee VS, Laub G, et al. Abdominal MR imaging with a
volumetric interpolated breath-hold examination.
Radiology1999; 212:876
-884[Abstract/Free Full Text]
- Bossuyt PM, Reitsma JB, Bruns DE. Towards complete and accurate
reporting of studies of diagnostic accuracy: the STARD initiative.
BMJ 2003;326:41
-44[Free Full Text]
- Connolly MM, Dawson PJ, Michelassi F, Moossa AR, Lowenstein F.
Survival in 1001 patients with carcinoma of the pancreas. Ann
Surg 1987;206:366
-373[Medline]
- Shimamoto K, Ishiguchi T, Sakumi S. CT evaluation of pancreatic
cancer: analysis of resected tumors. Eur J Radiol1987; 7:37
-41[Medline]
- Dupuy DE, Costello P, Ecker CP. Spiral CT of the pancreas.
Radiology1992; 183:815
-818[Abstract/Free Full Text]
- Vellet AD, Romano W, Bach DB, Passi RB, Taves DH, Munk PL.
Adenocarcinoma of the pancreatic ducts: comparative evaluation with CT and MR
imaging at 1.5T. Radiology1992; 183:87
-95[Abstract/Free Full Text]
- Gabata T, Matsui O, Kadoya M, et al. Small pancreatic
adenocarcinomas: efficacy of MR imaging with fat suppression and gadolinium
enhancement. Radiology1994; 193:683
-688[Abstract/Free Full Text]
- Semelka RC, Ascher SM. MR imaging of the pancreas.
Radiology1993; 188:593
-602[Abstract/Free Full Text]
- Kim MJ, Mitchell DG, Ito K, Kim PN. Hepatic MR imaging: comparison
of 2D and 3D gradient-echo techniques. Abdom Imaging2001; 26:269
-276[CrossRef][Medline]
- Sheridan MB, Ward J, Guthrie JA, et al. Dynamic contrast-enhanced
MR imaging of suspected pancreatic cancer: a comparative study with receiver
operating characteristic analysis. AJR1999; 173:583
-590[Abstract/Free Full Text]
- King AD, Ko GT, Yeung VT, Chow CC, Griffith J, Cockram CS. Dual
phase spiral CT In the detection of small insulinomas of the pancreas.
Br J Radiol1998; 71:20
-23[Abstract]
- Semelka RC, Custodio CM, Cem Balci N, Woosley JT. Neuroendocrine
tumors of the pancreas: spectrum of appearances on MRI. J Magn
Reson Imaging 2000;11:141
-148[CrossRef][Medline]
- Bluemke DA, Cameron JL, Hruban RH, et al. Potentially resectable
pancreatic adenocarcinoma: spiral CT assessment with surgical and pathologic
correlation. Radiology1995; 197:381
-385[Abstract/Free Full Text]
- Fletcher JG, Wiersema MJ, Farrell MA, et al. Pancreatic malignancy:
value of arterial, pancreatic, and hepatic phase imaging with multi-detector
row CT. Radiology2003; 229:81
-90[Abstract/Free Full Text]
- Johnson PT, Outwater EK. Pancreatic carcinoma versus chronic
pancreatitis: dynamic MR imaging. Radiology1999; 212:213
-218[Abstract/Free Full Text]
- Bronstein YL, Loyer EM, Kaur H, et al. Detection of pancreatic
tumors with multiphasic helical CT. AJR2004; 182:619
-623[Abstract/Free Full Text]
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
Top
Abstract
Introduction
