| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Gastrointestinal Imaging |
1 From the Departments of Radiology, Lucas MRS Center (R.W.P., L.C.C., C.F.B., M.N.M., R.E.M., R.B., R.B.J.) and Statistics (J.H.), Stanford University, Calif; and Department of Radiology, Veterans Administration Palo Alto Health Care System, Calif (M.N.M.). Received May 4, 2001; revision requested June 22; final revision received May 8, 2002; accepted May 28. R.W.P. supported by a research grant from the Max Kade Foundation. Address correspondence to R.W.P., Department of Radiology, University of Vienna, Währinger Gürtel 18-20, A-1090 Vienna, Austria (e-mail: rupert.prokesch@univie.ac.at).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Forty-three patients suspected of having pancreatic tumors underwent contrast material–enhanced biphasic multi–detector row computed tomography (CT). Curved planar reformations were generated along the pancreatic duct, common bile duct, and major mesenteric vessels. Three blinded independent readers assessed the curved planar reformations and transverse images separately for the presence of tumor, resectability, and vascular involvement. The results were compared with those of a consensus panel who evaluated the curved planar reformations and transverse images together along with clinical data and surgical findings.
RESULTS: Of 43 patients, 20 had pancreatic malignancies as judged by the consensus panel and proven at biopsy and/or clinical follow-up. For tumor detection, transverse images and curved planar reformations had an average sensitivity of 95.0% and 98.4% (P > .05), respectively, and an average specificity of 90.9% and 91.3% (P > .05), respectively. For tumor resectability, transverse images and curved planar reformations had an average sensitivity of 85.7% and 71.4% (P > .05), respectively, and an average specificity of 85.2% and 84.3% (P > .05), respectively. Average interpretation time was 6.4 minutes with transverse images and 4.1 minutes with curved planar reformations.
CONCLUSION: Curved planar reformations are equivalent to transverse images in the detection of pancreatic tumors and determination of surgical resectability.
© RSNA, 2002
Index terms: Computed tomography (CT), multi–detector row, 770.12119 • Images, analysis • Pancreas, CT, 770.12112, 770.12115, 770.12117, 770.12119 • Pancreas, neoplasms, 770.32
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
CT criteria for unresectability have included metastatic disease, most commonly involving the liver or peritoneum, occlusion of a major peripancreatic vein by tumor, encasement of a major peripancreatic artery, and contiguous invasion of adjacent organs such as the stomach and colon. Helical CT has allowed further refinements in determining the unresectability of pancreatic tumors due to better evaluation of vessel encasement. Additional CT criteria that have been evaluated include the degree of circumferential vascular involvement by tumor (9), the tear drop sign of the superior mesenteric vein (10), and the presence of dilated small peripancreatic veins, which are best seen on thin-section helical CT scans (11,12). At our institution, we have implemented additional curved planar reformations generated from transverse CT data for the evaluation of pancreatic tumors with the belief that they provide an excellent overview of the pancreas along its longitudinal axis. The purpose of this study was to evaluate the utility of curved planar reformations compared with standard transverse images in the assessment of pancreatic tumors.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
CT Imaging
CT scans were obtained with a multi–detector row CT scanner (LightSpeed QX/i; GE Medical Systems, Milwaukee, Wis) according to a dual-phase pancreatic protocol. An initial unenhanced localizer scan was obtained during a 10–12-second breath hold with 10-mm collimation, a pitch of 6 (high-speed mode), low milliamperage (80 mA), and 120 kVp. Immediately before scanning, the patient was asked to ingest 900–1,000 mL of water as a negative intraluminal contrast agent. A helical scan was then obtained that extended from 2 cm above the origin of the celiac trunk to 3 cm below the caudal extent of the pancreas. A small field of view (25 cm) was used, which was centered over the superior mesenteric artery. After an 18- or 20-gauge catheter (Angiocath; Becton Dickinson, Sandy, Utah) was placed into an antecubital vein, 140 mL of nonionic contrast material with an iodine content of 300 mg/mL (iohexol [Omnipaque]; Nycomed-Amersham, Princeton, NJ) was injected at 4 mL/sec by using a power injector (Envision CT; Medrad, Indianola, Pa).
The scanning parameters were as follows: 120 kVp, 200–240 mA, 4 x 1.25-mm collimation, rotation time of 0.8 second, pitch of 6 (high-speed mode), and scanning delay of 40 seconds (pancreatic phase). A subsequent portal venous phase scan was obtained from the diaphragm to the symphysis pubis after a scanning delay of 70 seconds. Scanning parameters included 120 kVp, 200–240 mA, 4 x 5-mm collimation, rotation time of 0.8 second, and pitch of 6 (high-speed mode). The matrix size was 512 x 512 in both series. The images obtained during the pancreatic phase were reconstructed at 0.5-mm intervals. These data were then transferred to an independent workstation.
Curved Planar Reformations
Curved planar reformations allow single two-dimensional image display of structures, such as airways and blood vessels, that run through multiple oblique planes. By using thin collimation and fast scanning techniques (ie, multi–detector row CT), even the pancreatic and common bile ducts can be depicted in their entirety. To create these images, we used two workstations (Advantage Windows 3.1, GE Medical Systems, Milwaukee, Wis; and Vitrea 2.2, Vital Images, Minneapolis, Minn). Modern versions of three-dimensional software are capable of automatically extracting median centerlines of vessels and creating curved planar reformations. Because pancreatic and common bile ducts are small in comparison to major blood vessels, this method has not yet been optimized for their display. Therefore, at present, creation of these images is totally operator dependent.
At our institution, full-time radiologic technologists who specialize in three-dimensional imaging and have been trained by radiologists routinely generate curved planar reformations and other three-dimensional reconstructions. These technologists are highly skilled and very efficient in this process, which truly is their area of expertise. In the event of complicated cases, a radiologist is always available for consultation. To generate these images, the operator must designate points on a two-dimensional image, following the course of the center of the pancreatic or the common bile duct. Lines through the volume connect these points, and the software extrapolates a two-dimensional image for display. The thickness of the curved plane is the voxel dimension perpendicular to the curved plane and depends on the orientation of the section on which it is drawn. The section thickness of the curved plane will never be larger than the effective section thickness or smaller than the transverse pixel dimensions. Two orthogonal curved planes are created through each structure, including a curved transverse, a curved sagittal, or a coronal reformation. The total time for obtaining the curved planar reformation is about 20 minutes. Because curved planar reformations are generated from two-dimensional images, creation of a three-dimensional volume is not necessary. Segmentation, or editing to mask out overlying structures, is also not required. This aids in the speed of creation of these images, making them practical in a clinical setting. Curved planar reformations were obtained of the pancreatic duct, the celiac artery, the hepatic artery, the splenic artery, the superior mesenteric artery, the portal vein, the superior mesenteric vein, and the splenic vein (Fig 1).
|
|
|
|
The consensus readers were presented with both curved planar reformations and transverse images during the same session and were asked to consider the information from both display modes in reaching a final interpretation of the case. In addition, these readers were given clinical data that included patient findings at presentation, age, and biopsy or surgical findings. With this additional information, we assumed that the readers would arrive at the most accurate overall interpretation of the imaging data. By implementing the consensus group, we could compare the results of the individual experimental readers with this independent standard, thereby enabling a determination of overall reader accuracy on each of the display modes. The mean number of transverse images to read was 421 (range, 277–567), whereas curved planar reformation consisted of a set of 14 images (range, 11–16).
Since we were interested in only imaging features that determined local resectability, sites of extrapancreatic spread such as liver lesions, regional lymph node enlargement, and peritoneal implants were not considered. Transverse images and curved planar reformations derived from arterial phase data were analyzed for the presence of tumor, tumor size, and tumor conspicuity. The tumor conspicuity was rated by using a five-point scale based on visible tumor-pancreas attenuation difference, secondary signs such as mass effect and ductal obstruction, and extrapancreatic extension: 1 = isoattenuating, no mass effect, only secondary signs; 2 = isoattenuating, mass effect or moderate attenuation difference without mass effect; 3 = moderate attenuation difference, mass effect or clear attenuation difference without mass effect; 4 = clear attenuation difference, mass effect; and 5 = clear attenuation difference, mass effect, obvious extrapancreatic extension. The pancreatic duct was analyzed for normal caliber (<3 mm), diffuse dilatation, or focal dilatation. The common bile duct was assessed for normal caliber (<7 mm) or dilatation.
The amount of suspected circumferential vessel involvement was recorded for the celiac artery, hepatic artery, splenic artery, superior mesenteric artery, portal vein, superior mesenteric vein, and splenic vein. Circumferential involvement by the tumor was categorized as follows: category 1, less than one quarter circumference of vessel contiguous with tumor; category 2, between one quarter and one half of vessel circumference contiguous with tumor; category 3, more than one half and up to three quarters of vessel circumference contiguous with tumor; and category 4, more than three quarters of vessel circumference contiguous with tumor (9).
Criteria for unresectability were vascular involvement (more than 50% circumferential encasement and/or occlusion) and extrapancreatic extension (excluding the duodenum). The confidence level for tumor resectability was rated from 1 to 5 and was evaluated in each case. The interpretation time was assessed separately for each display modality. CT findings were compared with findings at surgery and with the final pathology report, when available.
Statistical Analysis
A McNemar test for paired data was used for each reader separately to compare the sensitivities and specificities of transverse images and curved planar reformations with regard to tumor detection, resectability, and vascular involvement. The level of significance was indicated by a P value of less than .05. Sensitivity, specificity, positive predictive value, and negative predictive value were calculated for each modality and each reader separately.
Interobserver agreement among the three readers reading transverse images and curved planar reformations was quantified by using 
statistics. A value of less than 0.20 was considered to indicate poor agreement; a value of 0.20–0.39, fair agreement; a value of 0.40–0.59, moderate agreement; a value of 0.60–0.79, substantial agreement; and a value of 0.80 or greater, excellent agreement.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
Of the 43 patients, 20 were rated by the consensus panel as having positive findings for pancreatic cancer. The consensus readers considered seven tumors as resectable (Fig 2) and 13 as unresectable (Fig 3). Reader A rated 22 patients as having pancreatic cancer by using only curved planar reformations and 21 patients as having pancreatic cancer by using only transverse images. He rated 10 tumors as resectable and 12 as unresectable on curved planar reformations. On transverse images, he rated nine tumors as resectable and 13 as unresectable. Reader B rated 21 patients as having pancreatic cancer by using curved planar reformations and 25 patients as having pancreatic cancer by using only transverse images. He rated eight tumors as resectable and 13 as unresectable on curved planar reformations and 10 as resectable and 15 as unresectable on transverse images. Reader C rated 23 patients as having pancreatic cancer by using curved planar reformations and 20 patients as having pancreatic cancer by using only transverse images. Reader C considered 11 tumors resectable and 12 unresectable on curved planar reformations and 11 resectable and nine unresectable on transverse images.
|
|
|
|
|
|
|
|
|
|
|
values for tumor resectability were 0.67 and 0.64 for transverse images and curved planar reformations, respectively. For tumor detection, the values were 0.70 for transverse images and 0.65 for curved planar reformations. values for vessel involvement were 0.68 for transverse images and 0.66 for curved planar reformations. Generally, values ranging from 0.60 to 0.79 indicate substantial consistency. The average examination interpretation time for readers A-C was 6.4 minutes ± 2.17 for transverse images and 4.1 minutes ± 1.55 for curved planar reformations. The average examination interpretation time for the consensus panel was 13.5 minutes ± 4.38. By using a five-point scale, the average confidence level for tumor resectability was 4.3 ± 0.10 with transverse images and 4.2 ± 0.5 with curved planar reformations for readers A-C. In the consensus group, the average confidence level for tumor resectability was 4.8 ± 0.57.
Analysis of tumor size and the diameter of the common bile and pancreatic ducts revealed no difference between transverse images and curved planar reformations.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
One of the most common causes of unresectability not detected with CT is vascular involvement by tumor. The sensitivity of helical CT for vascular invasion is reported to be between 60% and 89% (13,14). In patients with pancreatic head tumors, staging of circumferential vessel involvement by tumor as shown with CT has proved to be a useful tool in predicting surgical resectability (9). Because survival may be increased only in patients in whom complete surgical resection with negative margins is possible, the aim of our study was to determine if curved planar reformations are a useful tool in the evaluation of pancreatic tumor resectability. Recent advances in multi–detector row CT technology have allowed the acquisition of near isotropic volumetric datasets. Postprocessing of these data with various three-dimensional reconstruction techniques such as curved planar reformation, volume rendering, and maximum intensity projection permits excellent display of anatomic information in multiple formats, which are likely to be more easily interpreted by referring physicians and surgeons (17,18).
Our data suggest that interpretation of curved planar reformations and transverse images results in no significant difference in detection of pancreatic tumors or determination of their resectability. While the data suggest that curved planar reformations are significantly equivalent to transverse images in the detection of pancreatic tumors and determination of vascular involvement of these tumors, it is important to point out that they do not represent a replacement for the transverse images. The interpreter of these images must realize that the curved planar reformation is only 1 pixel thick and does not represent the organ in its entirety. Additionally, there are many structures and portions of the anatomy that are not included on the curved planar reformations, and incidental findings in other organs would be easily overlooked if the curved planar reformation is evaluated in isolation. Thus, correlation with source images is imperative.
There were several limitations of this study. One possible bias involves the degree of familiarity that readers have with transverse images when compared with curved planar reformations. Readers are experienced in the interpretation of transverse images, have a systematic approach to reading such images, and are well versed in the distinction between normal and abnormal images. In contrast, curved planar reformations are a new addition to our protocol, with which readers have had substantially less experience. Additionally, curved planar reformation creates images in an arbitrary plane, which can be disorienting. While the structure of interest is exquisitely displayed, the neighboring anatomy becomes distorted. For these reasons, it would be expected for there to be a learning curve associated with the interpretation of the curved planar reformation. Unfortunately, this type of bias was unavoidable.
Ideally, the study protocol would have included evaluation of the curved planar reformations and transverse images together, as well as in isolation, by the blinded readers in comparison to the results of consensus panel. Unfortunately, this would have introduced additional memory bias, which could have skewed interpretation of the images. Additionally, because the order of the presentation of curved planar reformations and transverse images was not randomized, an order bias may have been introduced. Furthermore, a true pathologic standard was not available in all cases, and thus, a consensus panel that had access to all available clinical information was substituted.
While not an uncommon tumor, the small overall numbers of pancreatic malignancies seen in our institution resulted in a relatively small population size for our study. With only 20 cases, differences in the interpretation of presence or absence of tumor and resectability in just a few cases resulted in larger differences in individual reader’s sensitivity and specificity. The small sample size could also have resulted in a limitation of statistical power, preventing demonstration of a significant difference in detection of pancreatic tumors or determination of their resectability with the two display methods.
The most important clinical utility of curved planar reformations is most likely their ability to provide an excellent and quickly comprehensible overview of pertinent anatomy and structures for referring clinicians. This type of information can be of great use to surgeons in preoperative planning. With current multi–detector row CT technology and protocols, biphasic pancreatic studies easily produce hundreds of transverse images. While it remains necessary for the radiologist to evaluate each of the transverse images, the curved planar reformation provides a much more time-efficient and intuitive method of display for referring physicians interested in reviewing the pertinent findings. Finally, since our aim was to focus on tumor resectability, we did not perform a dedicated comparison of the performance of curved planar reformations versus transverse images in patients without tumors. It would be of interest in later studies to determine the value of curved planar reformation in nonmalignant pancreatic disorders.
In summary, in this cohort of patients, we were unable to detect a significant difference between curved planar reformations and transverse images for the detection and local staging of pancreatic carcinoma.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  K. Strobel, S. Heinrich, U. Bhure, J. Soyka, P. Veit-Haibach, B. C. Pestalozzi, P.-A. Clavien, and T. F. Hany Contrast-Enhanced 18F-FDG PET/CT: 1-Stop-Shop Imaging for Assessing the Resectability of Pancreatic Cancer J. Nucl. Med., September 1, 2008; 49(9): 1408 - 1413. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Furukawa, K. Uesaka, and N. Boku Treatment Decision Making in Pancreatic Adenocarcinoma: Multidisciplinary Team Discussion With Multidetector-Row Computed Tomography Arch Surg, March 1, 2008; 143(3): 275 - 280. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 G. A. Zamboni, J. B. Kruskal, C. M. Vollmer, J. Baptista, M. P. Callery, and V. D. Raptopoulos Pancreatic Adenocarcinoma: Value of Multidetector CT Angiography in Preoperative Evaluation Radiology, December 1, 2007; 245(3): 770 - 778. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Yanaga, K. Awai, Y. Nakayama, T. Nakaura, Y. Tamura, M. Hatemura, and Y. Yamashita Pancreas: Patient Body Weight tailored Contrast Material Injection Protocol versus Fixed Dose Protocol at Dynamic CT Radiology, November 1, 2007; 245(2): 475 - 482. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Ichikawa, S. M. Erturk, H. Sou, H. Nakajima, T. Tsukamoto, U. Motosugi, and T. Araki MDCT of pancreatic adenocarcinoma: optimal imaging phases and multiplanar reformatted imaging. Am. J. Roentgenol., December 1, 2006; 187(6): 1513 - 1520. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. Schueller, W. Schima, C. Schueller-Weidekamm, M. Weber, A. Stift, M. Gnant, and R. Prokesch Multidetector CT of Pancreas: Effects of Contrast Material Flow Rate and Individualized Scan Delay on Enhancement of Pancreas and Tumor Contrast Radiology, November 1, 2006; 241(2): 441 - 448. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
P. T. Johnson and E. K. Fishman IV Contrast Selection for MDCT: Current Thoughts and Practice Am. J. Roentgenol., February 1, 2006; 186(2): 406 - 415. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
E. K. Paulson, J. P. Harris, T. A. Jaffe, P. A. Haugan, and R. C. Nelson Acute Appendicitis: Added Diagnostic Value of Coronal Reformations from Isotropic Voxels at Multi-Detector Row CT Radiology, June 1, 2005; 235(3): 879 - 885. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. Imbriaco, A. J. Megibow, A. Ragozzino, R. Liuzzi, P. Mainenti, S. Bortone, L. Camera, and M. Salvatore Value of the Single-Phase Technique in MDCT Assessment of Pancreatic Tumors Am. J. Roentgenol., April 1, 2005; 184(4): 1111 - 1117. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Kobayashi, T. Yamaguchi, T. Ishihara, T. Ohshima, I. Ohno, K. Seza, Y. Shirai, K. Sudo, A. Nakagawa, H. Tadenuma, et al. Assessment of Portal Vein Invasion in Pancreatic Cancer by Fusion 3-Dimensional Ultrasonography J. Ultrasound Med., March 1, 2005; 24(3): 363 - 369. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
K. Takeshita, K. Kutomi, K. Takada, H. Kohtake, and S. Furui 3D Pancreatic Arteriography with MDCT During Intraarterial Infusion of Contrast Material in the Detection and Localization of Insulinomas Am. J. Roentgenol., March 1, 2005; 184(3): 852 - 854. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
