| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Orginal Research |
1 Department of Radiology, Soonchunhyang University Hospital, 657 Hannam-Dong,
Youngsan-Ku, Seoul 140-743, Korea.
2 Department of Radiology, College of Medicine, Ewha Womans University,
Tongdaemun Hospital, Jongro-Ku, Seoul, Korea.
3 Department of Applied Statistics, Daejeon University, Daejeon, Korea.
4 Department of Radiology, Cornell University Weill Medical College, New York,
NY.
Received October 19, 2006; accepted after revision March 28, 2007.
Address correspondence to J. H. Kim
(junghkim{at}hosp.sch.ac.kr).
OBJECTIVE. The objective of our study was to assess the diagnostic performance of virtual gastroscopy using MDCT for the detection of early gastric cancer (EGC) compared with 2D axial CT, focusing on interobserver reliability.
MATERIALS AND METHODS. During an 11-month period, we performed CT examinations of 94 consecutive patients with EGC and a control group composed of 68 patients without EGC. Three radiologists retrospectively interpreted the 2D axial CT and virtual gastroscopy images. Diagnostic performances were compared within each observer using the area under the receiver operating characteristic curve (Az). Sensitivity and specificity were also calculated for each individual observer. The simple kappa statistic was used to evaluate interobserver reliability in the detection of EGC.
RESULTS. The diagnostic performance for overall lesion detection in
patients with EGC was significantly higher with virtual gastroscopy
(Az = 0.829-0.885) than with 2D axial CT
(Az = 0.734-0.793) (p < 0.001). The
sensitivity and specificity of virtual gastroscopy for EGC were 78.7-84.0% and
83.8-91.2%, respectively. The sensitivity and specificity of 2D axial CT for
EGC were 62.8-69.2% and 83.8-92.7%, respectively. Virtual gastroscopy showed a
higher sensitivity for EGC than 2D axial CT (p < 0.001). The
interobserver reliabilities showed moderate to substantial agreement (
= 0.40-0.74 for 2D axial CT, = 0.57-0.71 for virtual gastroscopy).
CONCLUSION. Virtual gastroscopy showed excellent results with a good interobserver reliability for the detection of EGC compared with 2D axial CT.
Keywords: 2D CT • gastric cancer • MDCT • oncologic imaging • virtual gastroscopy
Gastric cancer is still one of the major causes of cancer deaths in the world, although mortality has been consistently decreasing. It is an aggressive tumor with a 5-year survival rate of less than 20%. However, early gastric cancer (EGC) is a curable disease, with a 5-year survival rate of more than 90%. Therefore, early detection of gastric cancer is essential [1-3]. EGC is defined as a tumor confined to the mucosa and submucosa regardless of nodal and distant metastases.
There is a geographic variation in the incidence of gastric cancer: The incidence is highest in Korea, Japan, China, South America, and Eastern Europe. As a consequence, screening of the asymptomatic population for gastric cancer has been implemented in Japan, Chile, and Venezuela. Improvement in survival has been most striking in Japan, where up to 40% of tumors are detected at an early stage due to vigorous screening programs [4-6].
Gastroscopy and upper gastrointestinal series are of great value for the detection of EGC. CT is currently the technique of choice for staging gastric cancer because it can be used not only to detect the primary tumor but also to predict the depth of invasion of the tumor and detect nodal involvement and distant metastases. However, the rate of EGC detection on 2D axial CT is low, between 20% and 56% [7-14].
MDCT scanners allow thinner and faster scanning that, in turn, results in excellent imaging resolution and easy generation of 3D image reconstructions. The use of 2D multiplanar reconstructions (MPRs) and virtual gastroscopy using a volume-rendering (VR) technique is a promising 3D imaging technique for the preoperative evaluation of gastric cancer. Application of these techniques for the detection of gastric cancer using MDCT has been reported [15-22].
To our knowledge, only a few reports regarding 3D imaging techniques for the detection of EGC have been published to date [14, 19]. Virtual gastroscopy provides gastroscopic viewing; therefore, it is expected to overcome the limitations of 2D axial CT in the detection of EGC. The purpose of this study was to assess the diagnostic performance of virtual gastroscopy using MDCT for the detection of EGC compared with 2D axial CT focusing on interobserver reliability.
Materials and Methods
Patients
This study was approved by our institutional review board for human
investigation, and informed consent was obtained from all patients. From May
2003 to April 2004, 94 consecutive patients with pathologically proven EGC
were enrolled in this study. All patients underwent gastroscopy. Patients
underwent MDCT for preoperative staging. During the same time period, 82
patients without EGC were referred to CT because of vague abdominal symptoms,
including epigastric pain and discomfort (n = 57), dyspepsia
(n = 12), nausea (n = 6), anorexia (n = 4), and
weight loss (n = 3). Of these 82 patients, 14 were excluded either
because they did not undergo gastroscopy (n = 5) or because they had
gastric lesions, including ectopic pancreas (n = 2), gastrointestinal
stromal tumor (n = 5), or erosive gastritis (n = 2).
The final study population consisted of 162 patients including a study group of 94 patients with EGC and a control group of 68 patients without EGC. The mean interval between gastroscopy and CT was 3.5 days (range, 0-7 days). The study and control groups included 95 men and 67 women with an age range of 35-73 years (mean, 56 years). EGC diagnosis was confirmed by means of surgery in 57 patients and endoscopic mucosal resections in 37 patients. The mean interval between CT and surgery or endoscopic mucosal resections was 9 days (range, 1-17 days).
CT Examination
For the detection of subtle gastric lesions, adequate distention is
essential. If the entire stomach is not well distended, disease may be
obscured or, conversely, the collapsed gastric wall may mimic disease.
Negative oral contrast medium with effervescent granules is more effective for
optimal gastric distention and for generation of virtual gastroscopy images
than contrast media without these granules.
At our institute, 6 g of effervescent granules (Top, Taejoon Pharmaceuticals) with a small amount of water was used to distend the patient's stomach. Before the patient underwent CT, 20 mg of scopolamine (Buscopan, Boehringer International) was injected intramuscularly to relax the bowel wall and reduce peristaltic bowel movement. A scanogram was obtained to ensure adequate gastric distention. CT was performed using an MDCT scanner (Sensation 4, Siemens Medical Solutions). The imaging parameters included a 4 x 2.5 mm section detector collimation, 120 kV, 145 mAs, 15 mm per second table speed, pitch of 6, and 1.5-mm reconstruction interval. Scanning covered the entire stomach region.
CT was first performed with the patient in the prone position; IV contrast enhancement was not used. Before the patient was scanned in the supine position, gastric distention was again checked with a second scout image; additional effervescent granules were given if the patient's stomach was not adequately distended. Because contrast enhancement facilitates better staging for gastric cancer and better evaluation of other abdominal organs, scanning with the patient in the supine position was performed after IV injection of 150 mL of ionic contrast material (Iopamiro 300 [iopamidol], Bracco; or Ultravist 370 [iopromide], Schering) at a rate of 3 mL/s through an 18-gauge angiographic catheter inserted into an antecubital vein. Scanning began 70 seconds after the initiation of the contrast injection, which corresponded to the portal venous phase. Scanning was performed from the diaphragmatic dome to the renal pedicle during a single breath-hold with the patient in the supine position.
Image Analysis
Two experienced gastrointestinal radiologists and one chief resident
retrospectively interpreted 2D axial CT images and virtual gastroscopy images
independently; the two gastrointestinal radiologists had 7 years of experience
in gastrointestinal radiology and experience interpreting virtual gastroscopy
images for more than 150 cases with correlation between findings at virtual
gastroscopy and results at gastroscopy. The chief resident had experience
interpreting virtual gastroscopy images for more than 60 cases with
correlation between findings at virtual gastroscopy and results at
gastroscopy. All observers were blinded to the gastroscopic results and to the
locations of EGC lesions.
Two-dimensional axial CT images were evaluated on PACS monitors in random order independently. The reconstructed image data sets were sent to the 3D workstation via the network. Virtual gastroscopy images were analyzed in random order independently on a commercially available workstation (Leonardo, Siemens Medical Solutions) using a software package with VR capabilities. The observers determined the presence or absence of a gastric tumor and, if present, its location. They rated the likelihood of EGC on a 5-point scale as follows: 1, no lesion; 2, lesion probably absent; 3, indeterminate; 4, lesion probably present; or 5, lesion definitely present.
A lesion was determined to be cancerous on 2D axial CT when the gastric wall showed focal thickening of 6 mm or greater or when focal enhancement was seen in the gastric wall with or without a low-density stripe at the base of the lesion corresponding to the submucosal layer. A lesion was determined to be cancerous on virtual gastroscopy when the gastric mucosa showed a polypoid lesion, a superficial lesion with elevation, a flat and depressed lesion, or an excavated lesion. A lesion also was determined to be cancerous on virtual gastroscopy when the gastric mucosa showed clubbing, abrupt cutting, and fusion of the converging folds at the margin of a depressed lesion. The location of each EGC lesion was classified as gastric cardia, fundus, body, angle, or antrum. The interval between review sessions was at least 2 weeks to eliminate learning effects. All observers recorded the time needed to interpret the images.
Statistical Analysis
For each observer, the diagnostic performance of the imaging procedures was
expressed by the area under the receiver operating characteristic (ROC) curve
(Az). Conventional measurements of sensitivity and specificity were
calculated using ratings of 1, 2, and 3 as absence of EGC and 4 and 5 as
presence of EGC. The McNemar test was used to compare the individual
accuracies of 2D axial CT and virtual gastroscopy. Interobserver reliability
was analyzed using the simple kappa statistic for a given imaging procedure
and in each possible pair of observers using the binary diagnosis of absence
or presence of EGC. The difference in interobserver reliability of 2D axial CT
and virtual gastroscopy was tested with a paired Student's t test
based on the two kappa values calculated in each possible pair of observers.
The degree of interobserver reliability, as indicated by kappa values, was
interpreted as follows: 0-0.20, slight agreement; 0.21-0.40, fair agreement;
0.41-0.60, moderate agreement; 0.61-0.80, substantial agreement; and 0.81-1,
almost perfect agreement.
A p value of less than 0.05 was considered to indicate statistical significance. ROC curve analysis was conducted using MedCalc software (version 4.20, MedCalc) for Windows (Microsoft), and other statistical analyses were conducted using SAS software (release 8.02, SAS Institute) for Windows.
Results
The macroscopic findings of EGC were classified according to the system established by the Japanese Research Society for Gastric Cancer: Type I is a polypoid lesion; type IIa, a superficial elevated lesion; type IIb, a superficial flat lesion; type IIc, a superficial depressed lesion; type III, an excavated lesion; and mixed type [23, 24]. According to the pathologic specimen and gastroscopy findings, EGC was classified as type I in seven patients, type IIa in 14 patients, type IIb in nine patients, type IIc in 27 patients, type IIa plus IIc in 27 patients, type IIc plus IIa in six patients, and type III in four patients. Lesions were located in the antrum in 50 patients, body in 30 patients, angle in 13 patients, and cardia in one patient.
|
|
|
= 0.40-0.74 on 2D axial CT, = 0.57-0.71 on virtual
gastroscopy). There was no statistically significant difference in
interobserver reliability of 2D axial CT and virtual gastroscopy (p =
0.27).
|
|
Nineteen EGC lesions that were not detected on 2D axial CT by at least one of the three observers were detected on virtual gastroscopy by all three observers. These 19 EGC lesions were morphologically classified as type I in one patient, type IIa in two patients, type IIb in one patient, type IIc in seven patients, type IIa plus IIc in seven patients, and type III in one patient (Fig. 2A, 2B, 2C). These lesions were located at the antrum in 12 patients, body in five patients, and angle in two patients.
|
|
|
|
|
|
|
|
Discussion
Gastroscopy and upper gastrointestinal series have been used for the detection of EGC. CT has been used for the evaluation of gastric cancer, mainly for staging, but not for detection. Virtual gastroscopy was recently introduced as a potential diagnostic tool for the detection of EGC. Virtual gastroscopy using MDCT showed excellent results in the detection of EGC. In our study, the diagnostic performance for overall lesion detection in patients with EGC was significantly higher with virtual gastroscopy (Az = 0.829-0.885) than with 2D axial CT (Az = 0.734-0.793). Moreover, in our study, interobserver reliability showed a moderate to substantial degree of agreement among the observers, which indicated a good diagnostic reproducibility of the technique. The results showed that virtual gastroscopy is an important tool in the detection of subtle mucosal lesions that cannot be detected on 2D axial CT.
According to previous reports for which investigators used a single-detector helical CT scanner, the rate of EGC detection on 2D axial CT was low, between 20% and 56% [7-14]. In patients with EGC, 2D axial CT images obtained using old-generation CT units usually did not show wall thickening. Ironically, if no abnormality was found on CT in patients with known gastric cancer, the cancer was considered to be EGC [9, 11-13]. In our study, the sensitivity for EGC on 2D axial CT using MDCT with thin sections improved markedly to 62-66%. Despite this technical improvement, 2D axial CT has limitations in detecting EGC lesions.
The use of 2D MPRs and of virtual gastroscopy with the VR technique is a promising method for the detection of EGC. Some investigators have reported the application of these techniques using MDCT for EGC detection [15-22]. According to those reports, the rate of EGC detection was higher using 2D MPR or virtual gastroscopy than using 2D axial CT alone, but the results were variable. Shimizu et al. [22] reported that the rate of EGC detection was 41.2% using 2D MPR with the water-filling method. According to that report, the detection rate for grossly nonprotruding EGC lesions, such as stage IIc, was relatively low (28.6%). These results indicate that 2D MPR with the water-filling method is not suitable for the detection of EGC.
Kim at al. [20] reported that the detection rate in 45 patients with EGC was 69% using 2D axial CT, 80% using 2D MPR, and 96% using virtual gastroscopy. According to that report, virtual gastroscopy has great value for the detection of EGC, which is similar to our results; however, our results showed a relatively lower sensitivity (78-84%) than previous reports [16, 17, 20]. We speculate that this lower sensitivity reflects differences in study design. In previous studies, investigators focused on preoperative staging of gastric cancers, so there was a potential bias in image interpretation. Although blinded to the results of the surgical or histopathologic analysis of the resected specimen, the observers were aware of the presence of gastric cancer. This knowledge probably affected the detection rate of EGC in some of these studies. However, our cohort included both a control group without gastric cancer and a study group with EGC, so the detection rates are lower than those cited in previous reports.
In our study, 19 newly detected EGC lesions seen in 15 patients on virtual gastroscopy were pathologically classified as type IIc, type IIa plus IIc, and type III. These results indicate that the strength of virtual gastroscopy lies in the detection of nonprotruding EGC lesions. Because gastric wall thickening is not seen in patients with these types of lesions, these EGC lesions may not be detected on 2D axial CT.
In our study, EGC lesions that were superficial and flat (type IIb) were missed in seven of nine patients and EGC lesions at the gastric angle were missed in seven of 13 patients. These missed cases may indicate that the weakness of virtual gastroscopy lies in the detection of superficial flat lesions (type IIb) and lesions at the gastric angle and that virtual gastroscopy has no advantage over 2D axial CT in the diagnosis of these types of EGC lesions.
Several limitations of virtual gastroscopy should be mentioned. First, it is a time-consuming examination to perform and interpret. Although greater computer processing power makes performing virtual gastroscopy faster, it still requires a long time to process. Moreover, because of the increased data volume, it requires more time to interpret. We reviewed the entire stomach on virtual gastroscopy using both the prone and supine data sets. When a tumor is in the antrum or body, the images obtained with the patient in the supine position are better than those obtained with the patient prone because the antrum and body are distended with air. On the other hand, when a tumor is in the cardia or fundus, the images obtained with the patient in the prone position are better because the cardia and fundus are distended with air. Second, virtual gastroscopy cannot be used to evaluate perigastric state or to detect distant organ metastasis. Therefore, virtual gastroscopy must be interpreted in conjunction with 2D MPRs for accurate staging of gastric cancer. Third, virtual gastroscopy cannot depict changes in the color of stomach mucosa. On gastroscopy, we can detect mucosal color changes and use chromogastroscopy during the examination [25]. Last, there is a long learning curve before one becomes proficient in the use of these tools. For detecting EGC on virtual gastroscopy, radiologists should be familiar with the appearances of EGC lesions and should be able to recognize characteristic morphologic changes and abnormal mucosal patterns.
Recent advances in CT technology and 3D imaging software enable us to obtain more detailed gastric images. Virtual gastroscopy is a promising method for detecting EGC despite its limitations. The advantages of virtual gastroscopy are that it has a wider field of view than conventional gastroscopy, the angle of the virtual camera can be adjusted omnidirectionally, and it has no blind point because retrospective reconstruction is available. In the near future, improvements of MDCT technology, 3D imaging software, and computer processing power will make faster and more accurate gastric imaging available.
We can use this technique for the early detection of gastric cancer in regions where gastric cancer is common. Mass screening for gastric cancer can be performed with virtual gastroscopy in regions where gastric cancer is common, although further studies are required to prove accuracy, efficacy, and cost-effectiveness of virtual gastroscopy as a screening method for EGC.
In conclusion, virtual gastroscopy showed excellent results in the detection of EGC compared with 2D axial CT, with a good interobserver reliability. Early detection and accurate preoperative staging of EGC offer the best prognosis and are essential for planning the optimal therapy, such as endoscopic mucosal resection or gastric resection. However, virtual gastroscopy is limited for detecting superficial flat lesions and some depressed lesions—especially those located in the gastric angle.
References
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
