| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1 Department of Diagnostic Radiology, Severance Hospital, and Research Institute
of Radiological Science, Yonsei University College of Medicine, Seodaemun-ku
Shinchondong 134, Seoul 120-752, Republic of Korea.
2 Brain Korea 21 Project for Medical Science, Yonsei University College of
Medicine, Seoul 120-752, Republic of Korea.
3 Department of Surgery, Yonsei University College of Medicine, Seoul 120-752,
Republic of Korea.
4 Department of Internal Medicine, Division of Gastroenterology, Yonsei
University College of Medicine, Seoul 120-752, Republic of Korea.
Received October 29, 2003;
accepted after revision December 28, 2003.
Supported by Yonsei University Research Fund of 2003.
Abstract
 Top Abstract IntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
SUBJECTS AND METHODS. Sixty-two patients with surgically proven rectal carcinomas underwent pelvic MRI before and after a rectal distention by warm water. Four radiologists, who were blinded to the study, reviewed each set of T1- and T2-weighted axial images obtained before and after the rectal distention and scored the image of the tumor. The accuracies for determining the tumor penetration of the outer wall of the rectum and the regional lymph node involvement were compared by analyzing the area under the receiver operating characteristic curve (Az).
RESULTS. For all reviewers, the tumor depiction scores were significantly higher in the distended images (3.8–3.9 for reviewers 1–4) than in the nondistended images (3.0–3.2) (p < 0.01). For determining the outer wall penetration, the accuracy of the three reviewers was significantly higher with the rectal distention images than with the nondistended images (p < 0.05). The mean accuracy of the all reviewers was significantly better with the distended images (Az = 0.922) than with the nondistended images (Az = 0.841) (p < 0.05). For determining the presence of regional lymph node involvement, all the reviewers came to similar conclusions in analyzing the two image sets.
CONCLUSION. Rectal distention by water filling may improve the depiction of a primary rectal tumor and the assessment accuracy of a perirectal tumor extension, but it does not improve the accuracy for determining the presence of regional lymph node involvement.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
The remaining 62 patients (42 men and 20 women; mean age, 56 years; age range, 28–80 years), who underwent radical surgery within 1 month (minimum, 1 day; maximum, 23 days; average, 8 days) of their MRI, were included in this study. None of the 62 patients underwent preoperative radiotherapy or chemoradiotherapy before surgery. Most patients underwent transrectal sonography performed by the same colorectal surgeon. However, the examination could not be completed in approximately one quarter of them because the tumors were located proximally or were stenotic or bulky. The sonographic data were not analyzed for the purpose of this study. The relative staging accuracy between transrectal sonography and MRI using an endorectal coil was evaluated previously in a blinded study in our institution; the accuracy was similar using the two imaging techniques [15]. All patients underwent curative surgery performed by the same surgeon. The surgical resection specimen was used as the gold standard for determining the depth of tumor penetration and the number of regional lymph nodes involved according to the recently revised TNM staging system [16] (Table 1). Pathologists with expertise in gastrointestinal pathology reviewed all materials.
|
MRI
MRI was performed using a 1.5-T system (Signa Horizon, General Electric
Medical Systems) and a phased array multicoil in all patients. The
examinations were performed in accordance with the requirements of our
institutional review board, and informed consent was obtained from all
patients. The patients did not undergo special bowel preparation before they
appeared in the MRI room. However, they were asked to defecate before the
examination if required.
After the localizer images were obtained, 20 mg of scopolamine butylbromide (Buscopan, Boehringer Ingelheim) was injected IV to minimize the peristalsis and alleviate the rectal spasms, if not contraindicated.
Nondistended axial T1-weighted images were obtained using a spin-echo sequence with the following parameters: TR range/TE range, 500–550/8–10; field of view, 24–26 cm; two signals averaged; matrix size, 256 x 192; and a section thickness of 5–6 mm with a gap of 1–2.5 mm. Both sagittal and axial T2-weighted images were obtained using a fast spin-echo sequence (4,000–6,000/75–105; echo-train length, 10–12; matrix size, 512 x 256) using the same field of view, scan thickness, and gap as used for the T1-weighted images.
Warm water was then administered using a balloon-tipped rectal tube, and the rectum was filled until the patient indicated a sensation of fullness in the rectum. The volume ranged from 200 to 500 mL. The rectal tube was removed after instilling the water.
The axial T1-weighted spin-echo images and axial and sagittal T2-weighted fast spin-echo images with the rectum distended were obtained using the same parameters as previously described.
Image Analysis
Four gastrointestinal radiologists independently and separately analyzed
the two image sets (i.e., MR images with the rectum nondistended and
distended). Both the nondistended and distended sets included the T1- and
T2-weighted axial images and the T2-weighted sagittal images. The reviewers
knew only that the patients had been referred for the preoperative staging of
rectal cancer and were unaware of the final surgical and histologic results.
The intervals between the reviews of the two image sets were at least 2 weeks.
To decrease the recall and learning bias from reviewing the same case, each
reviewer observed half of the nondistended images and half of the distended
images first and then observed the other half of the images at the second
session of the review. All the images were reviewed at a 1.5 x 1.5 K,
full PACS (picture archiving and communication system) workstation (Pathspeed
8.1, General Electric Medical Systems).
Each reviewer recorded scores for the depiction of the rectal mass on the basis of a 4-point rating: 1, tumor could not be identified; 2, tumor was identified, but its intraluminal margin was not defined; 3, presence of the tumor and its intraluminal margin was partially defined; and 4, entire tumor was well defined.
Each reviewer also recorded the confidence level for the presence of a perirectal extension, as defined by the outer wall penetration, and regional lymph node involvement using the following score for each item: 1, definitely absent; 2, probably absent; 3, possibly present; 4, probably present; and 5, definitely present. An interruption of the outer muscular layer, an irregular spiculation, and a grossly nodular appearance of the outer wall of the rectum were used as indicators of a perirectal extension. Regional lymph node involvement was deemed to be present when a perirectal lymph node larger than 0.5 cm in diameter was present or there was a perirectal lymph node with a heterogeneous signal intensity or an irregular border [17].
Finally, two reviewers jointly reviewed the nondistended and distended MR images separately and scored the depiction of the layers of the rectal wall in terms of the differentiation of the outer muscular layer and the inner submucosal layer by the following 4-point scale: 1, differentiation of the rectal wall layers was hardly perceptible; 2, differentiation of the layers was moderately observable in some areas of the rectum; 3, the differentiation of the layers was moderately possible in substantial areas of the rectum; and 4, the rectal wall layers were nicely differentiated in most areas of the rectum. Two reviewers discussed their findings and recorded the scores by consensus.
Statistical Analysis
The scores for depicting the rectal mass, as rated by each reviewer on the
nondistended and distended images, were compared using Wilcoxon's signed rank
test. To assess whether tumor depiction is affected by the tumor location
(< 5.0 cm, 5.1–10.0 cm, and > 10 cm from anal verge), the
morphologic type of the tumor (polypoid or fungating, ulcerative or
infiltrative, ulcerative and fungating), the tumor size (
3 cm,
3.1–6.0 cm, and > 6 cm), and the average of depiction scores of each
reviewer was compared by the Friedman analysis of variance test, which is
recommended to appropriately handle the reviewer effect of the ordinal
variables. The scores for depicting the rectal wall layers in the distended
and nondistended images were also compared by Wilcoxon's signed rank test.
The reviewer performance for determining an outer wall penetration and regional lymph node metastasis on the nondistended and distended sets was tested by a receiver operating characteristic (ROC) curve analysis. The true-positive fraction was plotted against the percentage of false-positive findings at each confidence level. The diagnostic accuracy was measured using the area under the ROC curve (Az). The Az values were compared to determine any differences in diagnostic performance using the Hanley and McNeil method [18]. To calculate the statistical significance of the difference between the averaged Az values, we used a multireviewer–multicase ROC approach by the jackknife method [19]. This method has the advantage of accounting for variability among techniques (in this case, nondistended vs distended rectal MR images), among the reviewers (in this case, four reviewers), and among the cases. The statistical significance of the results was reported at 95% confidence intervals (CIs) for the mean differences in Az values for reviewer performance using two image sets. A 95% CI that does not encompass zero is equivalent to a p value of less than 0.05 [20]. The mean Az values of all the observers for determining the presence of outer wall penetration were also calculated in each subgroup classified by tumor location, gross type, and size. Each reviewer's sensitivity, specificity, and accuracy were calculated for the diagnoses of perirectal extension and regional lymph node involvement, with the decision threshold scores from 3 to 5 being taken as positive for making a diagnosis in both cases. The interobserver agreement on the MR image interpretation was quantified using the kappa statistics with the following criteria: a kappa value of less than 0.20 indicated poor agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and 0.81 or greater, excellent agreement. The mean kappa values between each observer using each technique were compared using a paired Student's t test. A two-tailed p value of less than 0.05 was considered significant.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
|
|
|
|
Each reviewer's accuracies (Az values) for determining the presence of rectal wall penetration and regional lymph node metastasis are shown in Table 4. For determining the presence of outer wall penetration, the accuracy of the three reviewers was significantly higher using the rectal distention images than with the nondistended images (p < 0.05) (Figs. 2A,2B,3A,3B,4A,4B). The mean accuracy of all the reviewers was also significantly better with the distended images (Az = 0.924) than with the nondistended images (Az = 0.841) (95% CI: –0.1531 to –0.0109). The mean Az values of all the reviewers were higher using the distended MR images regardless of tumor location, gross type, and tumor size (Table 5). In terms of sensitivity and specificity, the distended images tended to improve the specificity more than the sensitivity for all the reviewers (Table 6). For determining the presence of a regional lymph node metastasis, two reviewers performed better on the distended images, whereas the other two performed better on the nondistended images, but no statistically significant difference was apparent for all reviewers.
|
|
|
|
|
|
|
|
|
The interobserver agreement for determining the presence of an outer wall
penetration was fair or substantial for the nondistended MR images (mean
± standard deviation, 
= 0.64 ± 0.07) and substantial or
excellent for the distended MR images ( = 0.78 ± 0.03)
(Table 7). The interobserver
agreement for determining the presence of regional lymph node involvement was
fair to substantial for both the nondistended and distended MR images. The
mean kappa values tended to be higher on the distended MR images for
determining the presence of both an outer wall penetration and regional lymph
node involvement.
|
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
Endorectal sonography is considered the most accurate technique for evaluating the local extent of rectal cancer. However, it is limited in evaluating stenosing or proximally located tumors, and the lymph nodes far from the rectal wall are difficult to evaluate [23]. Although the results of previous reports on MRI regarding the accuracy of the preoperative staging of rectal cancer vary [24–27], recent reports on high-spatial-resolution MRI using pelvic phased array coils have shown better results [2, 3]. To improve the staging accuracy, Wallengren et al. [3] introduced a technique involving the use of an enema containing a ferristene solution, with or without an IV gadodiamide injection. Others have also developed rectal distention techniques using superparamagnetic iron oxide solutions, diluted barium suspension, or methylcellulose [5, 7–9].
This study used warm water to distend the rectum. The water enema technique has been advocated by several authors for preoperative CT staging of colorectal carcinomas [28–30]. The rationale for this technique is that luminal distention of the rectum by water creates a clear distinction between the lumen and the inner wall of the rectum and helps to distinguish the tumor from the fecal material remaining in the rectum. A greater distention of the outer wall may also improve the demarcation between the tumor and the perirectal fat [29]. Rectal water also replaces any air remaining in the rectum, which may increase a susceptibility artifact during MRI. All patients tolerated this procedure and the motion artifacts could be eliminated by an injection of an antispasmodic agent.
Lupo et al. [31], who investigated the preoperative staging accuracy for rectal cancer using CT in subjects with or without a water enema, found that the water enema CT was significantly more accurate than the nonenema CT, with an accuracy of 84% versus 62% for the overall staging according to a modified Dukes' classification. Wallengren et al. [3] examined 29 patients using MRI with an enema of a ferristene solution plus the IV administration of gadodiamide and obtained a sensitivity of 100% and a specificity of 70% when distinguishing between the tumor stages worse than Dukes' stage A. Maier et al. [9] obtained a sensitivity of 97%, a specificity of 50%, and an accuracy of 82% for staging a rectal carcinoma higher than the T2 stage. However, to our knowledge, an intraindividual comparison between nondistended and distended MR images for the preoperative staging of rectal carcinomas has not yet been performed.
In this study, the MR images obtained after a rectal distention significantly improved the depiction of the primary tumors. Although the nondistended images failed to depict the rectal tumors in a few cases, these unidentified tumors could have been advanced stages that involved the perirectal fat and the lymph nodes. The accuracies of the preoperative staging were also improved in these cases by rectal distention.
In this study, the nondistended images also depicted most of the tumors, which is in contrast to the results reported by Wallengren et al. [3], in which the tumors were not depicted on the nondistended images in 25 (86%) of 29 patients. This difference can be attributed to the different study populations. In our study, only five (8%) patients were in stage T1, whereas 34% of the study population for Wallengren et al. was classified as Dukes' stage A. Furthermore, the use of relatively higher resolution fast spin-echo MRI in addition to the T1-weighted spin-echo images in our study may have improved the visibility of the tumor delineation on the nondistended images.
Tumor depiction with nondistended MRI was relatively poor for small (< 3
cm) or large (> 6 cm) tumors, tumors located in the distal rectum ( 5
cm from the anal verge), and polypoid-type tumors. This poor depiction was
probably caused by the fact that the distal rectum was collapsed in the
physiologic status; therefore, a small tumor that was usually polypoid was not
demarcated by surrounding fluid in the nondistended images. In addition, a
tumor that was too large was also poorly defined with the nondistended images.
However, tumor depiction with distended MR images did not vary significantly
according to location, gross type, and size of the tumors for either
technique. The superiority of the distended images was maintained for all
subgroups.
The differentiation of the rectal wall layers was better using distended MRI than with the nondistended images in a significantly larger number of patients. In the nondistended MR images, because the rectal wall is collapsed, irregular wrinkles can be seen along the inner and outer layer, and a rumpling of the inner linings makes it difficult to identify the layers clearly. However, in the distended MR images, the inner and outer layers are smoothed, and the difference in signal intensity was easier to define. Although the layers become somewhat thinned in most cases in the distended images, the differentiation of the layers was not affected much unless the lumen was overdistended, which was encountered in only a few cases.
In this study, rectal distention significantly improved the diagnosis of rectal wall penetration for three of the four reviewers, most significantly by increasing the specificity of a diagnosis of an outer wall penetration. One of the problems in the local staging of the rectal tumor was the difficulty in distinguishing perirectal inflammation and fibrosis from tumor infiltration. However, with luminal distention, the outer wall of the rectum at the site of the tumor was stretched and smooth, and the outer margin of the rectal wall was also less affected by the partial volume effect. Owing to these effects, the outer wall spiculation from a nontumorous origin observed on the nondistended images was less frequently observed on the distended images.
This study showed that warm water was also useful for improving local staging by MRI. However, this study could not address whether rectal distention using agents other than warm water, particularly more viscous agents, might have different results. Maier et al. [9] reported that high-viscosity ferristene formulations applied rectally were superior to low-viscosity formulations for the tumor staging. Recently, Goh et al. [7] used methylcellulose as a rectal contrast agent and suggested that a 5% weight and volume solution of methylcellulose would be appropriate for an adequate distention of the rectum. Although warm water was used in this study, the overall staging accuracy in both the nondistended and distended MR images was within the ranges reported by previous studies either with or without use of a rectal contrast medium [3, 6, 9, 32]. However, further studies will be necessary to determine the differential effect on the staging accuracy of different types of rectal contrast media.
The accuracy in determining the presence of regional lymph node involvement was comparable in the MR images obtained before and after rectal distention. This accuracy could be expected because rectal distention does not affect the depiction of the morphologic changes in the regional lymph nodes outside the rectal lumen. Although the rectal distention resulted in some displacement of the mesorectal fat, the vascular structures, and the lymph nodes, this displacement did not complicate the depiction of the nodal structures. According to a recent report by Brown et al. [17], the use of border contour and signal intensity characteristics instead of size criteria can improve the accuracy of nodal staging. Although our study used similar morphologic criteria for determining the presence of regional lymph node involvement, the accuracy in our study was not very high. Two reasons have been suggested. First, the inclusion of size criteria might have increased the false-positive rates. Second, the use of relatively thicker slice images made it difficult to evaluate the morphologic changes in the perirectal lymph nodes.
A limitation of our study is that it did not assess the utility of the rectal distention technique for evaluating the tumor relationship with the endopelvic fascia, internal or external sphincter, and puborectalis muscle. Recent reports have emphasized the role of preoperative MRI for evaluating the presence of tumor involvement of the circumferential resection margin or the anal sphincter [5, 6]. From the results of this study, we are not sure if a rectal distention is also helpful for evaluating the tumor involvement of these structures. However, a better delineation of the tumor, which was noted even in the T3 tumors in our series, may be helpful for estimating the exact tumor margin internally or externally.
Because the distended and nondistended MR images are obviously different, the reviewers could not be completely blinded in the assessment of the lesion depiction scores and the differentiation of the layers of the rectal wall. However, by separating the reviewing session in a long interval and by reviewing the images in random order, we tried to minimize the inherent bias in our study design.
The IV administration of a gadolinium-based contrast agent was not used in this study. There has been some controversy in using gadolinium-enhanced MR images for evaluating rectal carcinomas [3, 32–34]. In our experience, gadolinium enhancement may improve the tumor conspicuity on the T1-weighted images, but almost all lesions are also well depicted on the T2-weighted images after water distention. Gadolinium administration may erroneously accentuate the outer wall irregularity on the contrast-enhanced T1-weighted images by enhancing the perirectal vessels and inflammatory strands. This accentuation may increase the false-positive interpretation of the perirectal extension of the tumor [32, 34]. Gadolinium-enhanced images may be helpful for differentiating the tumor from fecal materials, but no substantial difficulty was apparent in distinguishing each of them by comparing the T1- and T2-weighted MR images, especially on the water-distended MR images.
Only the patients in whom the pathologic stages had already been proven surgically were included in this study. This limitation may have added a verification bias. Most of the patients, who were finally confirmed as having rectal carcinomas, were excluded from our study because they had advanced tumors. In these patients, the tumor penetration of the outer wall of the rectum and the regional lymph node enlargement were so obvious that the two imaging techniques showed the same results regarding the preoperative staging. Therefore, we believe that the exclusion of these patients would not have affected the comparative accuracy of the two techniques.
In conclusion, this study shows that a rectal distention by water filling may improve the depiction of tumors and the assessment accuracy of local tumor extension on preoperative MRI, but it did not significantly improve the assessment accuracy of the regional lymph node involvement. Further studies will be needed to determine the effect of the different types of rectal filling agent.
|
TopAbstractIntroductionSubjects and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  S. H. Kim, J. M. Lee, M. W. Lee, G. H. Kim, J. K. Han, and B. I. Choi Sonography Transmission Gel as Endorectal Contrast Agent for Tumor Visualization in Rectal Cancer Am. J. Roentgenol., July 1, 2008; 191(1): 186 - 189. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
X. M. Zhang, H. L. Zhang, D. Yu, Y. Dai, D. Bi, M. R. Prince, and C. Li 3-T MRI of Rectal Carcinoma: Preoperative Diagnosis, Staging, and Planning of Sphincter-Sparing Surgery Am. J. Roentgenol., May 1, 2008; 190(5): 1271 - 1278. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
H.-K. Chun, D. Choi, M. J. Kim, J. Lee, S. H. Yun, S. H. Kim, S. J. Lee, and C. K. Kim Preoperative staging of rectal cancer: comparison of 3-T high-field MRI and endorectal sonography. Am. J. Roentgenol., December 1, 2006; 187(6): 1557 - 1562. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
N. A. Obuchowski ROC Analysis Am. J. Roentgenol., February 1, 2005; 184(2): 364 - 372. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
