| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Gastrointestinal Imaging |
1 From the Department of Radiology, Juntendo University Urayasu Hospital, 2-1-1 Tomioka, Urayasu, Chiba 279-0021, Japan. From the 1998 RSNA scientific assembly. Received September 29, 1999; revision requested December 9; revision received February 3, 2000; accepted February 7. Address correspondence to N.S. (e-mail: shindoh@pop16.odn.ne.jp).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Upper gastrointestinal radiographs obtained within 3 years prior to diagnosis of gastric carcinoma in 336 patients were selected. Two radiologists who were initially blinded and then unblinded to the diagnosis reviewed the radiographs. Decisions were made by means of consensus. The reason for misdiagnosis was classified as perceptual error when the lesion was identified correctly at the blinded review, as possible perceptual error when the lesion was identified only at the unblinded review, and as technical error when the lesion could not be identified at either review and technical deficiencies were thought to be the cause.
RESULTS: Twenty-four patients underwent upper gastrointestinal radiography within 3 years prior to diagnosis of 27 carcinomas. The reason for misdiagnosis was classified as perceptual error in 11, as possible perceptual error in four, and as technical error in five lesions. In the remaining seven lesions, the lesion could not be identified at either review, and technical deficiencies were not thought to be the cause. The most common overlooked finding was depression (10 of 15), and the most common presumed technical error was incomplete compression study (seven of 11).
CONCLUSION: Careful attention should be paid to detect limited barium pooling during double-contrast studies to avoid overlooking depressions.
Index terms: Diagnostic radiology, observer performance, 70.123 • Endoscopy, 70.126 • Gastrointestinal tract, radiography, 70.123 • Radiography, comparative studies, 70.123 • Stomach, neoplasms, 70.321
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
At upper gastrointestinal radiography, accurate diagnosis of gastric carcinoma is an important goal for radiologists. It is necessary to have good knowledge of the perceptual and technical pitfalls of upper gastrointestinal radiography to improve the diagnostic accuracy for gastric carcinoma. In the literature (8–12), to our knowledge, little information is included on the pitfalls of biphasic upper gastrointestinal radiography in the diagnosis of gastric carcinoma. The purpose of this retrospective study was to evaluate the possible reasons for misdiagnosis of gastric carcinoma at routine upper gastrointestinal radiography and to assess how to reduce the occurrence of these errors.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
Previous upper gastrointestinal radiographs obtained in these patients were reviewed retrospectively twice by two experienced gastrointestinal radiologists (Y.O., S.K.), and decisions were made by means of consensus. At the first review, radiographs of 10 additional healthy subjects (six men, four women; age range, 42–65 years; mean age, 55.3 years) were included, and the radiologists were informed only that carcinoma was diagnosed in some patients within 3 years after the examinations. At the second review, the radiographs were reviewed in comparison with radiographs from which the diagnosis was made, with full knowledge of the endoscopic and histopathologic results. The radiologists tried at the second review to detect radiographic findings that reflected these final radiographic, endoscopic, and histopathologic results.
We classified the reasons for the initial misdiagnosis as follows: (a) perceptual error, when the lesion was correctly identified at the first review; (b) possible perceptual error, when the lesion was identified only at the second review; or (c) technical error, when the lesion could not be identified even at the second review and presumed technical deficiencies were present. When the lesion could not be identified even at the second review and presumed technical deficiencies were absent, we concluded that the lesion was at too early a stage for diagnosis.
Abnormal findings detected retrospectively were recorded at both reviews, and the presumed technical errors were recorded only at the second review. When multiple presumed technical errors were present, multiple errors were recorded. When presumed technical errors were present in cases of possible perceptual errors, the presumed technical errors were also recorded.
Macroscopic type, depth of invasion, and main site and shape—fundus, corpus, and antrum along the long axis and anterior wall, posterior wall, lesser curvature, greater curvature, and circularity along the short axis—of the lesion were recorded on the basis of histopathologic records of the resected specimen. When some part of the lesion was located on the lesser or greater curvature, except for circular lesions, the location was defined as located on the lesser or greater curvature. In patients whose lesions could not be resected, the macroscopic type and main site of the lesion were recorded by using upper gastrointestinal radiographic and endoscopic images. Macroscopic types were determined according to the classification of the Japan Gastroenterological Endoscopy Society for early carcinoma and the Borrmann classification for advanced carcinoma (Table 1) (12).
|
|
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
All patients underwent upper gastrointestinal radiography and endoscopy with biopsy at the time of diagnosis. The diagnosis of gastric carcinoma was made initially at upper gastrointestinal radiography in 20 patients with 21 lesions. In the remaining four patients with six lesions, the diagnosis was made initially at endoscopy, and upper gastrointestinal radiography also depicted these lesions after endoscopy. All lesions were detected at both final upper gastrointestinal radiography and endoscopy. In three patients, surgery was not performed because of carcinomatous peritonitis. In another patient, only gastrojejunostomy was performed because severe invasion of the pancreas and common bile duct was identified at laparotomy.
On the radiographs obtained in the 10 healthy subjects, no abnormal findings were recorded. The characteristics of the lesions in the 24 patients are summarized in Table 3. Seven of 27 lesions were classified as being at too early a stage for diagnosis because they could not be identified at either review and technical deficiencies were not thought to be the cause. These seven lesions consisted of three lesions of type IIc early carcinomas limited to the mucosa and four lesions of type 4 advanced carcinomas (scirrhous type) at diagnosis. In the remaining 20 of 27 lesions, considerable errors were identified. These 20 lesions consisted of 14 early carcinomas and six advanced carcinomas at diagnosis. The time between the two examinations in the 20 lesions ranged from 5 to 34 months (mean, 19.3 months).
|
The causes of errors are listed in Table 4. The reason for misdiagnosis was classified as perceptual error in 11 of 20 lesions. In seven of 11 lesions, a shallow depression was identified on double-contrast images (Fig 1). In two of 11 lesions, a shallow depression adjacent to the pylorus or duodenal loop was identified on compression images (Fig 2). In one of 11 lesions, radiating folds were identified with superimposed cardia on double-contrast images (Fig 3). In one of 11 lesions, a distinct double contour (tumor) was identified on double-contrast images (Fig 4). Among the 11 lesions, malignant appearance was identified in eight lesions: irregular margin of depression in four, irregular margin of depression with irregular mucosal nodularity adjacent to the depression in three, and tumor in one.
|
|
|
|
|
|
|
|
|
|
|
In the remaining five of 20 lesions, the reason for misdiagnosis was classified as technical error. Seven presumed technical errors were recorded in the five lesions. In two of five lesions, the region of interest—the greater curvature of the antrum—was not depicted because of superimposition of barium-containing duodenum on the double-contrast images and lack of a compression study. In one of five lesions, the barium coating was extremely poor in the region of interest (the lesser curvature of the proximal corpus). In two of five lesions, a compression study was lacking in the region of interest (the anterior wall of the antrum). In nine lesions, 11 technical errors were recorded.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
In previous studies (8–10), causes of error at upper gastrointestinal radiography were analyzed in comparison with endoscopic images or specimen examination. In this study, we analyzed the causes in comparison with the radiograph from which the diagnosis was made. We believe that comparison by means of the same diagnostic method is best for analyzing the cause of error in detail.
In this study, many overlooked lesions (15 of 20) were detected in retrospect. Therefore, it is important to recognize overlooked findings and the presumed reasons for their having been overlooked. In seven lesions, shallow depression was overlooked on double-contrast images. These lesions had shallow depressions with limited barium pooling, and accompanying radiating folds were absent or limited. In these depressions, a margin of depression with or without irregular mucosal nodularity adjacent to the depression was seen in retrospect. Careful reading of limited barium pooling is necessary to differentiate depressive lesions from normal residual barium on double-contrast images.
From a technical point of view, adequate gaseous distention may improve the depiction of such shallow depressions. Maruyama and Baba (12) reported that overdistention of the stomach may obscure shallow depressive lesions, and they recommended a two-step double-contrast study performed by using 2 g of effervescent granules in the first step and an additional 2–3 g of effervescent granules in the second step for routine upper gastrointestinal examinations. Their method may improve the depiction of the margin of depression and/or irregular mucosal nodularity adjacent to shallow depressive lesions.
Misinterpretation of abnormal findings as normal architecture is another presumed cause for misdiagnosis. In our series, radiating folds on double-contrast images were probably misinterpreted as normal cardia, and two depressions on compression images were probably misinterpreted as normal pylorus and superimposed duodenum. It is important to have a good knowledge of the radiographic appearance of normal architecture.
Overlooking the double contour is the remaining cause of misdiagnosis. Careful attention should be paid in reading the lesser and greater curvature. When extra opacities other than the normal lesser and greater curvature line are suspected, the possibility of a lesion should be considered (9). From a technical point of view, compression studies and steep oblique double-contrast studies may improve depiction of the lesion (Fig 4).
In nine lesions, 11 technical errors were recorded. The most common technical error was related to compression studies (seven of 11). In these lesions, the location was the anterior wall of the antrum in two, the greater curvature of the antrum in three, the lesser curvature of the distal corpus in one, and the lesser curvature of the antrum in one. Many lesions in the anterior wall of the distal half of the stomach can be depicted only by means of compression studies (13). Compression studies are also useful in lesions that are located in the lesser or greater curvature, because part of the lesion in the anterior wall can be depicted clearly.
An adequate compression study performed with the patient in the upright position is sometimes difficult in obese patients (12), and the usefulness of compression studies performed with the patient in the prone position was reported (11). Our routine examination did not include a compression study with the patient in the prone position. Compression studies with the patient in the prone position should be performed when compression studies with the patient in the upright position fail or are insufficient. Even when compression studies with the patient in the prone position fail, a double-contrast study with the patient in the prone right posterior oblique Trendelenburg position should be considered (11).
Another technical error was failure to separate the region of interest from the barium-containing duodenum in two lesions located in the greater curvature of the antrum. Superimposition of the barium-containing duodenum sometimes causes poor depiction of the distal corpus and antrum (11,13). To prevent superimposition of the barium-containing duodenum, authors of previous reports (11,13) have indicated the usefulness of obtaining initial double-contrast radiographs before spillage of barium into the duodenum. This method failed to depict the two lesions because of poor barium coating. Although steep right and left anterior oblique double-contrast spot radiographs were included in our routine examination, the area of interest of the two lesions failed to appear separate from the barium-containing duodenum because of an insufficiently oblique view.
The remaining technical error was poor barium coating in two lesions located in the lesser curvature of the proximal corpus, where poor barium coating sometimes occurs (12). In such conditions, additional turning of the patient in the Trendelenburg position can improve barium coating in this area.
In this retrospective review, more than half of the overlooked lesions were identified correctly by two experienced gastrointestinal radiologists without comparison with the radiographs from which the diagnosis was made, and there were no false-positive errors in healthy subjects. Our double reading of upper gastrointestinal radiographs usually was performed by a resident or fellow and an experienced gastrointestinal radiologist. Double reading by two experienced gastrointestinal radiologists may reduce the occurrence of perceptual errors.
In the category of perceptual errors, malignant appearance was identified retrospectively in eight lesions. In contrast, no malignant findings were identified in the other three lesions misdiagnosed owing to perceptual errors, but the findings were different from the typical erosions. Early carcinomas sometimes have a radiographic appearance similar to that of benign lesions (12). Therefore, endoscopy with biopsy should be performed in patients with radiographically abnormal findings other than typical erosions.
Several limitations of this study should be addressed. First, the initial review of radiographs was performed with the knowledge that the carcinoma was diagnosed within 3 years after the examinations in some patients. This condition may have increased the rate of perceptual errors. Second, all nondetected lesions that were in the anterior wall of the antrum and had inadequate compression studies were classified as technical errors. Some lesions may have been difficult to identify even if adequate compression studies had been performed, and these lesions should have been classified as being at too early a stage for diagnosis. This condition may have increased the rate of technical error.
In summary, many overlooked lesions were identified in retrospect. The most common perceptual error was overlooking of shallow depressions. Careful attention should be paid to detect limited barium pooling during double-contrast studies. The most common technical error was insufficient compression study. When compression studies performed with the patient in the upright position fail, other patient positions should be used.
|
FOOTNOTES |
|---|
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTSDISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  C.-Y. Chen, D.-C. Wu, Y.-T. Kuo, C.-H. Lee, T.-S. Jaw, W.-Y. Kang, and J.-S. Hsu MDCT for Differentiation of Category T1 and T2 Malignant Lesions from Benign Gastric Ulcers Am. J. Roentgenol., June 1, 2008; 190(6): 1505 - 1511. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
