| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Radiology, Ewha Womans University Mok Dong Hospital, 911-1
Mok-Dong, Yang-Cheon-Ku, Seoul, 158-710 South Korea.
2
Department of Radiology, Duke University Medical Center, Erwin Rd., Box 3808,
Durham, NC 27710.
3
Department of Radiology, Beth Israel Deaconess Medical Center, 330 Brookline
Ave., Boston, MA 02215.
Received October 2, 2000;
accepted after revision December 4, 2000.
Address correspondence to S. Y. Baek.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. For 44 patients with known pancreatic malignancy a total of 56 arterial phase helical CT scans were obtained. Targeted pancreatic imaging was performed, and reformatted images were generated. Axial source images, reformatted images, and the combination of axial and reformatted images were interpreted independently by three observers. The observers graded the celiac axis, common and proper hepatic, splenic, gastroduodenal, and superior mesenteric arteries for tumor involvement. Grades of vascular involvement were compared by intra- and interobserver variability analyses.
RESULTS. Intraobserver agreement averaged over five vessels was good
between the axial and combined techniques for each individual observer (0.64
0.66), but intraobserver agreement was poor between the
axial and reformatted ( = 0.17 and = 0.31, respectively) and
the reformatted and combined techniques ( = 0.31 and = 0.38,
respectively) for two observers. For grading of vascular involvement in each
vessel, intraobserver agreement was good to excellent between the axial and
combined techniques (0.48 0.82). Interobserver agreement
averaged over five vessels was poor for imaging techniques except between
observer 2 and observer 3 on the axial ( = 0.47) and combined
techniques ( = 0.47). For grading of vascular involvement in each
vessel, interobserver agreement for reformatted technique was poor (0.09
0.40).
CONCLUSION. Multiplanar and volume-rendered techniques showed the highest intra- and interobserver variability in grading vascular involvement by pancreatic malignancy. These images should be used in combination with routine axial images to decrease observer variability.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The role and utility of CT angiography in patients with known or suspected pancreatic malignancy continue, however, to be controversial. Raptopoulos et al. [4] reported that helical CT with CT angiography provided useful information about local vascular involvement from pancreatic carcinoma. On the other hand, Diehl et al. [1] reported CT angiography could not show all the findings seen on axial helical scans.
The purpose of our study was to compare reformatted (2D multiplanar and 3D volume rendered) images with the axial source images obtained during the arterial phase and with the combination of axial and reformatted images in the assessment of the arterial involvement by pancreatic malignancy. To the best of our knowledge, no study has evaluated this series of images in terms of intra- and interobserver variability analyses.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
A review of the histopathology reports revealed that 20 patients had pathologically proven pancreatic malignancy by biopsy or fine-needle aspiration of the pancreas, 14 patients by surgery, six patients by biopsy of hepatic metastases, one by biopsy of the omentum, one by biopsy of a lymph node along the superior mesenteric artery, one by cytology of bile duct brushing, and one by cytology of pancreatic duct brushing. Histopathologically, 43 patients had ductal adenocarcinomas and one patient had mucinous cystadenocarcinoma. Pancreatic malignancy was primarily located in the head in 34 patients, in the body in 13 patients, in the tail in one patient, and in more than one portion in eight patients.
Scanning and Reconstruction Techniques
All dual-phase helical CT examinations of the pancreas were performed on
HiSpeed Advantage scanners (General Electric Medical Systems, Milwaukee, WI).
For 35 examinations, 450 mL of a 1.2% barium sulfate suspension (Readi-CAT;
E-Z-Em, Westbury, NY) was administered, and for 21 examinations 400-500 mL of
water was routinely administered 20-30 min before the study to distend the
stomach, duodenum, and proximal jejunum. All patients received 175 mL of
iopamidol (Isovue 300 [300 mg/mL]; Bracco Diagnostics, Princeton, NJ) IV
injected at a rate of 4 mL/sec. After an initial delay of 20 sec, arterial
phase images were obtained from 1 cm above the celiac axis through the entire
pancreas. Scanning parameters included 140 kVp, 160-190 mA, a 1:1 pitch, 3-mm
collimation, 1-mm reconstruction interval, and small field of view (22-25
cm).
After the section data were reconstructed, all the information was transferred to the image server where it could be accessed by a dedicated workstation. All 2D multiplanar and 3D volume-rendered images were then generated using software (Vitrea; Vital Images, Minneapolis, MN) and a processing unit (O2; Silicon Graphics, Mountain View, CA).
A comprehensive systematic four-step approach was used to evaluate each volumetric data set. In step one, the axial images were surveyed using a cine-loop format for the presence and extent of lesion, particularly soft-tissue masses. In step two, a survey of the anatomy and integrity of major arterial branches was performed by rotating 3D volume-rendered "lighted" images. In step three, these 3D images were cropped electronically to isolate specific arteries (selective arteriography), particularly when there was vascular superimposition. In step four, 2D multiplanar images of the perivascular soft tissue were generated using curved projections in the coronal or sagittal plane. These images are critical for depicting the relationship of blood vessels to perivascular softtissue masses, particularly tumor involvement or lymph nodes.
Image Review
The axial sources images from the arterial phase, reformatted images (2D
multiplanar and 3D volume-rendered), and the combination of axial and
reformatted images (axial, 2D multiplanar, and 3D volume rendered) were
interpreted independently by three observers. The length of time between the
same observer's review of the same case with different techniques was 1 week.
Observer 1, observer 2, and observer 3 had 5 years, 8 years, and 3 years of
experience in abdominal CT, respectively, at the time of the study, but they
had almost the same amount of experience in interpreting reformatted imaging,
1-2 months in practice. The observers knew that each patient had a pancreatic
malignancy, but they were unaware of the patient's identity and clinical
history and the surgical findings, histopathologic findings, or both. The
observers were asked to score vascular involvement using a six-point scale as
follows: grade 0, normal, which was defined as fat or normal pancreas between
the tumor and the vessel; grade 1, loss of the fat between the tumor and blood
vessel without displacement of the vessel; grade 2, displacement or narrowing
of the vessel or both by soft tissue on one side or involving less than half
the diameter of the vessel; grade 3, encasement of the vessel by soft tissue
on both sides or more than half the vessel diameter; grade 4, vessel occluded
by a soft-tissue mass; and grade 5, vessel not seen. Involvement of five
peripancreatic arteries (celiac axis, common and proper hepatic arteries,
splenic artery, gastroduodenal artery, and superior mesenteric artery) was
graded on this six-point scale.
Grades of vascular involvement between imaging techniques scored by three
observers were compared to evaluate the intra- and interobserver variability
using the kappa measure of agreement (excellent, 
0.75; good, 0.40
< 0.75; poor, < 0.40). Agreement meant the exact
same score on the six-point scale.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
0.66) and between all imaging
techniques for observer 2 (0.48 0.64)
(Table 1).
|
Intraobserver agreement for grading of all vessels was good and excellent
between the combined and axial images for individual observers (0.48
0.82) (Table 2 and
Fig.
1A,1B,1C).
Especially for grading the celiac axis, there was excellent agreement between
the combined and axial images for observer 2 ( = 0.82) and observer 3
( = 0.80). Although intraobserver agreement for all vessels was good or
excellent between all imaging techniques for observer 2 (0.39
0.82), there was poor agreement for all vessels between the axial and
reformatted images for observer 1 and observer 3 (0.02 0.46)
(Table 2 and Fig.
1A,1B,1C).
|
|
|
|
The mean kappa measure of agreement between observer 1 and observer 2 for
all imaging techniques averaged over five vessels was 0.31, that between
observer 2 and observer 3 was 0.42, and that between observer 3 and observer 1
was 0.27 (Table 3). Therefore,
interobserver agreement was poor for imaging techniques except between
observer 2 and observer 3 on the axial images ( = 0.47) and combined
images ( = 0.47).
|
Interobserver agreement for all vessels was poor on reformatted images
(0.09 0.40) (Table
4 and Fig.
2A,2B,2C),
although interobserver agreement for the splenic artery was good on the axial
(0.42 0.49) and combined images (0.49
0.66) and there was good agreement for the celiac axis between observer 2 and
observer 3 on all imaging techniques (0.40 0.55) and between
observer 1 and observer 2 on axial and combined images ( = 0.40,
= 0.49, respectively).
|
|
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
With dual-phase helical CT it is possible to obtain images during the phase of maximum arterial enhancement and during the phase of optimal liver enhancement. The pancreas has a rich arterial supply and early scanning is helpful for evaluation of peripancreatic arterial involvement by the tumor [5]. Volumetric helical data sets provide not only axial sections but also multiplanar displays. Both 2D and 3D reconstructions are well established in imaging various regions [6, 7] and are especially useful for visualizing vascular structures.
The most widely used techniques for 3D imaging are shaded surface display, maximum intensity projection, and volume rendering [8]. CT angiography with volume rendering has advantages over shaded surface display and maximum intensity projection [8, 9]. Relative voxel attenuation is represented in 3D through a gray scale or in color, and because one can choose which voxels to render opaque and which to render transparent, there is more information than on either a shaded surface display or a maximum intensity projection. Volume rendering maintains the anatomic spatial relationship of the CT data set and depth information, thereby better representing the interrelationship between vascular and visceral structures [10]. We used volume rendering with surface shading to give the reviewers realistic 3D images.
Arterial and venous involvement by pancreas malignancy is important to recognize because it may render resection impossible [4]. The inability of CT to reveal vascular involvement accurately may be because of insufficient vascular enhancement or diminished resolution of cross-sectional imaging along the length of the vessels, such as the superior mesenteric vein, portal vein, and superior mesenteric artery [4]. The use of helical CT and CT angiography may help solve this problem by increasing vascular opacification and providing a more optimal format for vascular display [4]. Angiographic display improves the depiction of perivascular soft tissue and the conspicuity of subtle alteration in vascular caliber not detected on axial images alone [4].
Three-dimensional volume-rendered imaging of peripancreatic arteries has the potential for replacing the invasive angiographic technique [11]. It is fast and noninvasive and provides cross-sectional information about blood vessels and the surrounding soft tissue [12,13,14]. The role of 3D reconstructed images in determining vascular involvement in patients with a pancreas malignancy is controversial. Some authors have reported that 3D reconstructed vascular images provide useful information [4, 15], whereas other authors report that these images are of little value [1]. Still other authors have shown that 3D reconstructed vascular images are a useful adjunct to 2D axial images from dual-phase helical CT, particularly in patients with venous invasion by periampullary tumors [16] and pancreas adenocarcinoma [15].
Because we obtained 3D reformatted images of peripancreatic arteries during the arterial phase of helical CT, 3D reformatted venous images were not obtained. The degree of arterial contiguity by soft tissue is relevant for determining resectability. Any amount of arterial contiguity means the patient is incurable. Resection, however, may improve longterm survival. The criteria for unresectability for the arteries differ from that for the veins. Arterial contiguity may indicate unresectability. For the veins, there must be invasion of the wall or intraluminal tumor.
Noninvasive CT angiography has some limitations. It requires a large amount and high injection rate of IV contrast material, which may be associated with an adverse event, such as renal insufficiency or an extravasation injury. Nonfilling of veins, especially the inferior vena cava and portal vein, may also result in diagnostic difficulties. Because of the thin collimation, the number of axial images that must be obtained increases, requiring additional radiation exposure and increases in time and expense [4].
The five peripancreatic arteries (celiac axis, common and proper hepatic arteries, splenic artery, gastroduodenal artery, and superior mesenteric artery) are vulnerable to invasion by the tumor. Whether these arteries are encased by the tumor is important to determine resectability. Therefore, we interpreted the grade of the five arteries, and it was practical to decide the grade of five arteries independently.
A six-point scale was used to grade the vascular involvement in our study. The observers and surgeon considered grades 3, encasement of the vessel by soft tissue on both sides or more than half the vessel diameter, and above as unresectable. If we had interpreted images using a simpler scale, the variability might have been improved, but we considered the six-point scale to be accurate for interpreting the grade of vascular involvement.
The main objective of our study was to compare the different techniques for the intra- and interobserver variability in grading arterial involvement by tumor, not to determine the resectability of the tumor. Therefore, we graded each vessel carefully and our attention to each vessel was not diminished even though one vessel was involved with tumor. Observer variability reflects a difference in interpretation and does not always indicate a difference in diagnostic accuracy.
In our study, intraobserver agreement when averaged over five vessels was
good between the combined and axial techniques (0.64 0.66),
but was poor between the axial and reformatted techniques (0.17
0.48). Because the reformatted technique showed the highest intraobserver
variability, we believe that reformatted technique should not be used alone
(Table 1). Although the axial
and combined techniques revealed good agreement averaged over five vessels
( = 0.47) between observer 2 and observer 3, there was poor agreement
between the three observers on all imaging techniques. This means
interobserver agreement was poor for grading peripancreatic arterial
involvement (Table 3).
Grading of tumor involvement of the celiac axis had excellent agreement
between axial and combined images for observer 2 ( = 0.82) and observer
3 ( = 0.80). Intraobserver agreement was good and excellent for the
combined and axial techniques, respectively, for the three observers (0.48
0.82) (Table
2). Although interobserver agreement was good for grading the
splenic artery on axial and combined images (0.42 0.66),
agreement was poor for all vessels between the three observers on reformatted
images (Table 4). Therefore,
reformatted techniques revealed the highest intra- and interobserver
variability in grading arterial involvement by pancreatic malignancy.
The reasons why the reformatted techniques showed so much variability were as follows. First, the observers were not familiar with non-axial image display and had 1-2 months of experience interpreting the reformatted images before the study. Second, the axial source images were obtained routinely according to the CT protocol, so the observers interpreted the grade of vascular involvement without difficulty; however, the reformatted images were made to reveal the best several images showing vascular involvement, so the grade of vascular involvement was more difficult to interpret. Third, because the reformatted images were reconstructed with axial source images, the quality of reformatted images might be lower than that of axial source images because of artifacts.
The quality of the reformatted images might be improved with the use of multidetector CT technology. This improvement in quality would undoubtedly result in a better data set because thinner sections would be acquired with more anatomic coverage in less time (i.e., shorter breath-hold).
We tried to observe the axial source images as being equal to the reformatted images when we observed the combined techniques. Therefore, the variability between the axial and combined images could not have been less on the basis of the observer merely relying on the axial source images of the combined techniques.
Although the sensitivity of CT for predicting the unresectability of pancreatic carcinoma has approached 100%, the overall accuracy of CT for tumor staging is only 66-78% [5, 17,18,19]. The main limitations of CT are its failure to reveal involvement of great vessels by tumor and small hepatic and peritoneal implants [4]. Bluemke et al. [5] found unresectable disease due to undetected liver and peritoneal implants in 21% of patients considered to have resectable disease, and Raptopoulos et al. [4] found unresectable disease due to metastatic disease in 14% of patients who were considered to have resectable tumor. In our 14 patients who underwent surgery, six (43%) who were considered to have resectable tumors on CT were subsequently found to have undetected liver metastases at laparotomy.
More than half of the scans (35/56) in our study were obtained with an oral contrast agent by mistake. This might have caused the reformatted images to be degraded by artifact. Therefore, we evaluated the quality of the reformatted images with three grades: grade 1, no artifact; grade 2, moderate artifact but it is possible to interpret image; and grade 3, severe artifact that makes image difficult to interpret.
Observer 1, observer 2, and observer 3 interpreted the images for four cases, two cases, and three cases as grade 3, respectively. Therefore, 52-54 (93-96%) of 56 cases had images that were not affected by artifact. The main cause of artifact was stairstep artifact due to respiratory motion. Oral contrast agent hardly affected the quality of the reformatted images, especially the curved multiplanar reformatted images, because overlapped and oral contrast-filled bowel loops could be cropped electronically.
The main limitation of our study is that the six grades of peripancreatic arterial involvement could not be correlated with surgical and pathologic results. Surgery was performed in 14 patients. Of these 14 cases, only four were resectable and the other 10 were unresectable. In two resectable cases, all three observers interpreted CT as showing that resection was possible. However, in one of these cases, two observers interpreted images as showing involvement of the superior mesenteric artery, and in another case, the third observer interpreted images as showing involvement of the common hepatic and superior mesenteric arteries (Fig. 3A,3B). In six unresectable cases, small metastatic nodules were noted on the liver capsule and no further surgery was performed. Three cases revealed portal vein involvement, and one case revealed involvement of the superior mesentery artery. In the latter case, CT predicted involvement of superior mesenteric artery before surgery. Surgery was not performed in the other 42 cases because CT findings revealed unresectable pancreatic malignancy before surgery. This shortcoming, however, is not unique to our study mainly because it is difficult and perhaps not in the patients' best interest to obtain strict surgical correlation. Few studies have one-to-one surgical confirmation of arterial involvement and perhaps for good reason. If surgeons find a peritoneal implant, they will not pursue a vigorous anatomic dissection to evaluate each vessel. Furthermore, doing so might significantly and needlessly increase morbidity.
|
|
In conclusion, reformatted (2D multiplanar and 3D volume rendered) images showed the highest intra- and interobserver variability in grading arterial involvement by pancreatic malignancy. Although these techniques may be helpful in determining the presence or absence of vascular involvement in certain cases, they should be used in combination with the routine axial data set to minimize observer variability.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  E. K. Fishman, D. R. Ney, D. G. Heath, F. M. Corl, K. M. Horton, and P. T. Johnson Volume rendering versus maximum intensity projection in CT angiography: what works best, when, and why. RadioGraphics, May 1, 2006; 26(3): 905 - 922. [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]
|
|
|
|
|
|
D. D. Cody AAPM/RSNA Physics Tutorial for Residents: Topics in CT: Image Processing in CT RadioGraphics, September 1, 2002; 22(5): 1255 - 1268. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
