| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Clinical Radiology, Kyushu University Graduate School of Medical
Sciences, 3-1-1, Maidashi, Higashi-ku Fukuoka 812-8582, Japan.
2 Department of Surgery and Oncology, Kyushu University Graduate School of
Medical Sciences, Higashi-ku Fukuoka 812-8582, Japan.
3 Department of Anatomic Pathology, Kyushu University Graduate School of Medical
Sciences, Higashi-ku Fukuoka 812-8582, Japan.
Received October 23, 2001;
accepted after revision January 29, 2002.
Address correspondence to K. Yoshimitsu.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Two radiologists retrospectively reviewed both hard-copy and soft-copy (on a monitor with multiplanar reconstruction capability) versions of helical CT scans (3-mm collimation and 3-mm reconstruction) of 21 patients who had undergone surgical resection for carcinomas of the gallbladder. The local spread of the disease was evaluated according to the TNM system, and the results were correlated to the pathologic findings. Inter- and intraobserver differences were checked with kappa statistics. Results of the consensus interpretations were used to calculate sensitivity, specificity, and accuracy of helical CT.
RESULTS. No significant inter- or intraobserver differences were found in any T category evaluation. The sensitivities of the hard-copy consensus interpretations in the diagnosis of T1, T2, T3, and T4 lesions were 33%, 64%, 80%, and 100%, respectively; specificities of hard-copy interpretations were 94%, 80%, 81%, and 95%, respectively. For soft-copy (monitor) consensus interpretations, the sensitivities for the diagnosis of T1, T2, T3, and T4 lesions were 33%, 73%, 80%, and 100%, respectively; the specificities of soft-copy interpretations were 94%, 80%, 88%, and 95%, respectively. Overall accuracy of the hard-copy interpretation was 83%; the overall accuracy of the soft-copy interpretation was not significantly different—86%.
CONCLUSION. Helical CT provided 83-86% accuracy in the diagnosis of the local extent of carcinomas of the gallbladder, showing acceptable sensitivity and specificity for the T2 and more advanced lesions but poor sensitivity for the T1 lesions. Use of a monitor with multiplanar reconstructions of the CT data did not significantly improve the diagnostic accuracy.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The recent development of a helical CT scanning technique performed with bolus injection of a contrast medium with thinslice collimation has permitted better spatial resolution in the cephalocaudal plane in patients with pancreatic carcinomas [5,6,7]. To our knowledge, no report to date has been published regarding the performance of helical CT in the preoperative evaluation of local spread of gallbladder carcinomas. In this report, we present our findings for 21 patients with gallbladder carcinomas, all of whom underwent surgical resection. The purposes of this report were to assess the diagnostic accuracy of helical CT in staging tumors according to the TNM system [8] using interpretations of axial hard-copy scans and to evaluate whether interpretation of the images on a monitor with multiplanar reconstruction (MPR) capability improved diagnostic accuracy.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Helical CT Protocol
Patients fasted at least for 4 hr before the examination; no oral contrast
medium or water was given to the patients. All 21 patients were examined with
the same type of scanner (X Vigor; Toshiba, Tokyo, Japan); four phases of
scanning were performed. After an initial unenhanced 10-mm contiguous scan of
the upper abdomen was obtained, helical CT of the whole gallbladder was
performed with 3-mm collimation, 1:1 pitch, 200 mAs, and 120 kV. Scanning was
begun 45 sec after IV injection of a contrast medium containing 300 mg I/mL
(Omnipaque 300; Daiichi Pharmaceutical, Tokyo, Japan) with a total volume of
150 mL through a superficial venous system of the upper extremities at a rate
of 2-3 mL/sec. The patients were asked to hold their breath during the 30-sec
scanning time. The scanning covered 9 cm, typically beginning at the level of
the umbilical portion of the left portal vein and proceeding in a
cephalocaudal direction to the level of the root of the superior mesenteric
artery.
After allowing the patients to breathe for 10 sec, we performed helical scanning of the whole upper abdomen from the top of the liver to the lower pole of the kidneys to obtain portal venous phase images (net delay time, 85 sec). The parameters for this scanning were 7-mm collimation, 1:1 pitch, 150 mAs, and 120 kV. Equilibrium phase images were obtained 6 min after the IV injection of contrast medium using 7-mm collimation. Image reconstruction was performed at 3 mm for the second phase and at 7 mm for the third and fourth phases of the scanning.
Image Interpretation
The hard copies of the CT scans obtained with 3-mm collimation were
retrospectively and independently interpreted by two experienced abdominal
radiologists who were unaware of the surgical results. Although the reviewers
knew that all patients had confirmed carcinomas of the gallbladder, they were
unaware of tumor location and size. First, the reviewers were asked to locate
the lesions. They then were asked to evaluate the extent of the spread of the
disease using the TNM rating system
[8] after the precise lesion
location was revealed to them by the study coordinator.
Using previous pathology studies on gallbladder carcinomas
[9,10,11]
and the TNM system, we defined the CT findings for each T category as follows:
T1, polypoid lesions without focal thickening of the gallbladder wall; T2,
nodular or sessile lesions associated with focal thickening of the gallbladder
wall at what was considered to be attachment sites and with the presence of an
apparently smooth fat plane separating the adjacent organs; T3, lesions
showing loss of a fat plane separating the lesions from a single adjacent
organ, indicating tumor involvement (
2 cm into the liver) or with apparent
nodularity on the serosal aspect, indicating serosal exposure of the tumor;
and T4, lesions involving two or more adjacent organs or extending into the
liver more than 2 cm.
The reviewers rated the confidence level of their diagnosis of a lesion using a 5-point scale (1, definitely negative; 2, probably negative; 3, equivocal; 4, probably positive; 5, definitely positive) for each T category from T1 to T4. For example, if a reviewer interpreted a lesion to be probably at T2 but perhaps at T3, that reviewer may have rated the lesion as 1 for T1, 4 for T2, 3 for T3, and 1 for T4. If a reviewer interpreted a lesion to be definitely in the T4 category, then that reviewer may have given ratings of 1 for T1, 1 for T2, 1 for T3, and 5 for T4. In interpreting a lesion to be in the T3 or T4 category, the reviewers indicated the presence or absence of tumor infiltration into the liver, duodenum, colon, and bile duct.
After an interval of 8 to 9 weeks, the same two reviewers were asked to independently interpret the CT data of the same patient group on the operating console of the CT scanner. On this console, sagittal and coronal MPR images of the scanning performed with 3-mm collimination were displayed as well as the axial reference scans. Oblique MPR images were arbitrarily created on the console for three patients by one reviewer and for one patient by the other reviewer to optimize the images of the tumors. The images were reconstructed using a linear interpolation with 1.3-mm (minimal value available) to 3-mm thickness and were displayed on the console continuously with a manual track-ball. A 5-point scale indicating the extent of disease spread was applied in a similar fashion as that used for the hard-copy interpretations.
Data Analysis
A binomial receiver operating characteristic (ROC) curve was fitted to each
observer's confidence rating by an estimation of maximal likelihood
[12,
13]. The area under the ROC
curve (Az value) was evaluated for diagnostic accuracy. The
differences between Az values were tested by the CORROC2
[12,
13] software algorithm, using
the paired two-tailed t test. For analysis of interobserver or
intraobserver variability, kappa statistics were applied. Kappa values greater
than 0 were considered to be indicative of a positive correlation. Values 0.40
or less were considered to be indicative of a positive but poor correlation;
0.41-0.75, good correlation; and greater than 0.75, excellent correlation. The
sensitivity, specificity, and positive and negative predictive values were
calculated by defining confidence level ratings of 4 and 5 as positive and 3
or less as negative for both hard-copy and soft-copy interpretations:
discrepancies regarding the positivity or negativity of ratings between the
two observers were resolved by consensus. The sensitivity, specificity,
accuracy, and positive and negative predictive values of the hard-copy and
soft-copy interpretations were compared using Wilcoxon's signed rank test.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Tumor Detection
Of the 21 lesions, 19 were correctly located by both observers. Two lesions
with involvement of the neck or cystic duct of the gallbladder and with
mass-forming xanthogranulomatous cholecystitis at the fundi were not diagnosed
by either of the reviewers, and the xanthogranulomas were misinterpreted as
tumors.
Hard-Copy Interpretation of Lesion Spread
The data for the three T1 lesions were subjected to ROC analysis, and an
interobserver kappa value of 0.56 was obtained. Only one lesion was correctly
diagnosed; two were misdiagnosed as T2 lesions. The sensitivity, specificity,
and positive and negative predictive values of T1 lesions at consensus
interpretation were 33%, 94%, 50%, and 89%, respectively
(Table 1). Images from a
representative patient are shown in Figure
1A,1B,1C.
|
|
|
|
For T2 lesions, ROC analysis was applied, and the Az values of the two observers were 0.67 and 0.76, with no statistically significant difference between them (p = 0.45). An interobserver kappa value of 0.46 was obtained. The sensitivity, specificity, and positive and negative predictive values of T2 lesions at the consensus interpretation were 64%, 80%, 78%, and 67%, respectively (Table 1). Three T2 lesions—all of which had a small but apparent nodularity along their serosal aspect—were misinterpreted as T3 lesions. Close pathologic evaluation of the specimens revealed small, focally engorged vessels, possibly representing venous branches along the serosal aspect of the tumor (Fig. 2A,2B). In one of these three lesions, loss of the fatty plane of the gallbladder fossa, which may merely have been due to the general paucity of fat in this thin patient, was also overestimated as tumor invasion into the liver. In another T2 lesion, subserosal infiltration of the tumor was subtle histopathologically, which is what we believe led to the erroneous CT interpretation of it as a T1 lesion.
|
|
The data for T3 and T4 lesions were subjected to ROC analysis, and interobserver kappa values of 0.47 and 0.82 were obtained. The sensitivity, specificity, and positive and negative predictive values of T3 lesions at consensus interpretation were 80%, 81%, 57%, and 93%, respectively: those of the T4 lesions were 100%, 95%, 67%, and 100%, respectively (Table 1). Four T3 lesions were correctly diagnosed because of liver invasion in three patients and common bile duct invasion in one patient. In one of the T3 lesions with liver involvement, concurrent inflammatory change involving the adjacent omentum and the duodenum was misinterpreted as tumor infiltration (T4) by both observers. Two T4 lesions with hepatic and duodenal involvement were correctly diagnosed by both observers. Images from representative patients are shown in Figures 3A,3B and 4A,4B,4C.
|
|
|
|
|
The overall sensitivity, specificity, and accuracy of consensus hard-copy interpretations were 67%, 89%, and 83%, respectively (Table 1).
Soft-Copy Interpretation of Lesion Spread
ROC analysis to test the interobserver difference was available only at the
level of T2 evaluation. Data were insufficient for the evaluations of T1, T3,
and T4: the Az values of the two observers were 0.71 and 0.77,
showing no significant interobserver difference (p = 0.63). At each T
category, kappa values between the two observers ranged from 0.47 to 0.82,
showing good to excellent agreement.
At consensus interpretation, one of the T1 lesions that had been incorrectly diagnosed as T2 on the hard-copy scan was correctly diagnosed on the soft-copy version. However, another of the T1 lesions was erroneously diagnosed as a T2 lesion on the soft-copy version; that lesion was correctly categorized as T1 on the hard-copy scan. One of the T2 lesions that had been erroneously diagnosed as T3 on the hard-copy scan was correctly diagnosed at soft-copy interpretation. Overall sensitivity, specificity, and accuracy at the consensus soft-copy interpretation were 71%, 89%, and 86%, respectively (Table 1).
Comparison Between Hard-Copy and Soft-Copy Interpretations
Inter- and intraobserver differences, as described by kappa values, are
summarized in Table 2. All data
showed good to excellent agreement.
|
ROC analysis of intraobserver differences was available for only the T2 evaluation (Fig. 5), and data were insufficient for the evaluation of T1, T3, and T4 lesions. No significant intraobserver differences were found in the Az values of the hard-copy and soft-copy interpretations of either observer (p = 0.71 and 0.94, respectively). No significant difference was found in the diagnostic accuracy at consensus interpretation between the hard-copy and soft-copy interpretations in any T category for either observer (p > 0.999 in any T category, Wilcoxon's signed rank test). Overall sensitivity, specificity, and positive and negative predictive values were not significantly different between the hard-copy and soft-copy interpretations (p > 0.999, Wilcoxon's signed rank test).
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Our data suggest that a correct diagnosis of T1 lesions is often difficult. In this T category, we found poor sensitivity (33%) and positive predictive values (50%) but acceptable specificity (94%) and negative predictive values (89%). In our experience, the normal mucosal and muscular layers enhance as one unit or as one complex, probably because of the absence of a submucosal layer and the subserosal layer per se is hardly recognizable on helical CT data obtained using 3-mm collimination [14]. We therefore applied morphological criteria that were based on the previously reported pathologic observations [9,10,11] that most T1 lesions are polypoid with thin stalks and typically are 2 cm or less in diameter, whereas T2 lesions tend to be nodular or sessile with focal thickening of the adjacent gallbladder wall. Distinguishing T1 lesions from T2 lesions on helical CT using these criteria, however, was discouraging in our series, perhaps because of the limited spatial resolution achieveable with the current technique. If new technology is developed that allows the subserosal layer of the gallbladder wall to be consistently identified, new criteria for distinguishing between T1 and T2 lesions on CT may be found.
In contrast, the findings of our study show that the diagnosis of T2 or more advanced lesions appears more promising. Sensitivity, specificity, and positive and negative predictive values were all acceptable, ranging from 60% to 100%. The value of helical CT is therefore considered to lie in permitting differentiation of lesions confined within the serosa from those extending beyond the serosa.
Few reports have described the performance of CT in the assessment of the spread of gall-bladder carcinoma, perhaps because the rarity of the disease prohibits use of a consistent methodology in examining patients and because patients typically present when the disease is at an advanced stage, precluding the possibility of surgical confirmation. Ohtani et al. [4] reported on the performance of conventional CT; in that study, the researchers used three different CT systems and inconsistent scanning techniques and had discouraging results. Our results (overall accuracy of 83-86%) suggest that helical CT can provide better T staging of gallbladder carcinoma than conventional CT.
Histopathologic correlation revealed that in three of the T2 lesions, small vessels along the serosal aspect of the gallbladder wall caused nodularity along the serosal aspect of the tumor, resulting in the observers' overestimation (serosal exposure or T3) of the tumor spread. This nodularity may represent focally engorged cholecystic venous branches attributable to the increased blood flow by the tumor. Close evaluation of the vascular structure along the serosal aspect of the organ may prevent this misinterpretation. Development of techniques for achieving higher spatial resolution in images and more rapid delivery of the bolus injection of contrast medium (e.g., using multidetector CT [15]) may help solve these problems. In one patient in our study, a T3 lesion was marked by concurrent inflammation around the gallbladder, which mimicked peritoneal implants: in this patient, the lesion was erroneously diagnosed as a T4 lesion. Radiologists also need to be aware that lesions can be overstaged in the presence of concurrent inflammation.
Two patients in our study each had a small carcinoma that involved the neck of the organ exclusively. In both patients, the gallbladder had extensive xanthogranulomatous lesions at the fundi, and these latter lesions were erroneously interpreted as malignant. Several investigators [16,17,18] have attempted to differentiate this rare, benign entity from carcinomas of the gallbladder. Chun et al. [18] reported that the presence of intramural low-density nodules may be a sign of xanthogranulomatous cholecystitis. However, the xanthogranulomas in our patients did not appear to be such nodules and were misinterpreted as carcinomas, and the small malignant lesions at the neck were overlooked. The elevation of the intraluminal pressure and the resultant bile leak into the wall caused by the obstructive processes at the neck have been suggested to be among the causes of xanthogranulomatous cholecystitis [19]. The two cases of xanthogranulomas in our series may have represented an indirect sign of the carcinoma at the neck. Radiologists need to pay full attention to the entirety of the gall-bladder, particularly at the neck and cystic duct, even when a massive lesion is apparent at the fundus.
One of the limitations of this study is that MPR was performed without overlap, which resulted in suboptimal spatial resolution in the cephalocaudal direction on sagittal and coronal MPR images (despite application of linear interpolation during the reconstruction process). The lack of adequate spatial resolution may in part explain our finding that interpretation of MPR images with axial reference scans did not significantly improve diagnostic accuracy. Overlapping reconstructions of the images might have provided better diagnostic performance for interpretation of MPR images than was achieved in our study [20, 21]. Other limitations of this study include its retrospective nature, the small number of patients (particularly those with T1 and T4 lesions), and a possible bias in the patient population toward those with less advanced disease because we recruited patients with surgically confirmed lesions.
In conclusion, helical CT with 3-mm collimation and reconstructions provided 83-86% accuracy in the evaluation of the local spread (T factor) of the gallbladder carcinomas. Its usefulness was limited as a diagnostic aid in identifying T1 lesions but was acceptable in identifying T2 and more advanced lesions. Interpretation of soft-copy CT data with MPR and without overlapping reconstructions did not significantly improve the accuracy of diagnosis in our study.
Acknowledgments
We thank Masazumi Tsuneyoshi of the Department of Anatomic Pathology,
Kyushu University Graduate School of Medical Sciences, for providing the
pathologic specimens for this study.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. J. Kim, J. M. Lee, J. Y. Lee, J. Y. Choi, S. H. Kim, J. K. Han, and B. I. Choi Accuracy of Preoperative T-Staging of Gallbladder Carcinoma Using MDCT Am. J. Roentgenol., January 1, 2008; 190(1): 74 - 80. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. Kalra, S. Suri, R. Gupta, S. K. Natarajan, N. Khandelwal, J. D. Wig, and K. Joshi MDCT in the Staging of Gallbladder Carcinoma. Am. J. Roentgenol., March 1, 2006; 186(3): 758 - 762. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
D. Grand, K. M. Horton, and E. Fishman CT of the Gallbladder: Spectrum of Disease Am. J. Roentgenol., July 1, 2004; 183(1): 163 - 170. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) and soft-copy (
) interpretations of observer 1 when
making T2 determination. Values for area under ROC curve of hard-copy and
soft-copy interpretations were 0.71 and 0.67, respectively, showing no
statistically significant difference (p = 0.45).