| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Original Research |
1 Department of Surgery, Samsung Medical Center, Sungkyunkwan University School
of Medicine, Seoul, South Korea.
2 Department of Radiology and Center for Imaging Science, Samsung Medical
Center, Sungkyunkwan University School of Medicine, 50 Ilwon-dong, Kangnam-ku,
Seoul 135-710, South Korea.
Received July 16, 2005;
accepted after revision November 3, 2005.
Supported by a Samsung Medical Center Clinical Research Development Program
grant.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. During an 8-month period, 24 patients with rectal cancer underwent both 3-T MRI performed with phased-array coils and 7.5- to 10-MHz endorectal sonography in the 3 weeks before surgical resection. Three radiologists independently reviewed the MR and endorectal sonographic images. The histopathologic findings in resected specimens were used to evaluate the sensitivities and specificities of these techniques for invasion of the muscularis propria and perirectal tissue and for lymph node involvement. Receiver operating characteristic (ROC) analysis was used to compare the diagnostic accuracies of the techniques.
RESULTS. For muscularis propria invasion, the mean sensitivities of both MRI and endorectal sonography were 100%, and the mean specificities were 66.7% and 61.1%, respectively. The differences in the mean sensitivities and specificities were not statistically significant (p > 0.05 in each case). For perirectal tissue invasion, MRI and endorectal sonography had comparable sensitivities and specificities (91.1% vs 100%, 92.6% vs 81.5%; p > 0.05 in each case). They also had similar sensitivities and specificities for lymph node involvement (63.6% vs 57.6%, 92.3% vs 82.1%; p > 0.05 in each case). ROC curves for muscularis propria invasion and lymph node involvement showed no differences in diagnostic accuracy. The mean area under the ROC curve for endorectal sonography (Az = 0.996) for perirectal tissue invasion, however, showed higher accuracy than that of MRI (Az = 0.938, p = 0.028).
CONCLUSION. The sensitivity, specificity, and accuracy of 3-T MRI were similar to those of endorectal sonography for muscularis propria invasion and lymph node involvement, but for perirectal tissue invasion, 3-T MRI was less accurate than endorectal sonography.
Keywords: colon • MRI • rectal cancer • sonography
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MRI Techniques
All patients underwent MRI with a 3.0-T whole-body system (Intera Achieva
3T, Philips Medical Systems) with a dedicated cardiac sensitivity-encoding
coil (six elements of phased-array coil) at 80 mT/m and a slew rate of 200
T/m/s. No IV antiperistaltic agent was administered. The patients were asked
to perform rectal cleansing 2 hours before the MRI examination with two rectal
suppositories (bisacodyl, Dulcolax suppository, Boeringer Ingelheim). In the
MRI suite, warm water was administered with a balloon-tipped rectal tube, and
the rectum was filled until the patient reported it felt full. The volume of
water ranged from 150 to 400 mL. The rectal tube was removed after
instillation.
A coronal localizing image was obtained for selection of axial and sagittal images with a T2-weighted turbo spin-echo (TSE) sequence (TR/TE, 2,500-5,000/100; echo-train length of 6; 5-mm slice thickness; 1-mm gap; 256 x 256 matrix; 24-cm field of view, 2 signals acquired; sensitivity-encoding factor of 2; and sequence duration of 3-5 minutes). The sagittal images were used to plan thin-slice axial and coronal imaging. The axial and coronal T2-weighted TSE sequence (2,500-5,000/100; echo-train length of 6; 3-mm slice thickness; 1 mm gap; 312 x 312 matrix; 18-cm field of view; 4 signals acquired; sensitivity-encoding factor of 2; 1-mm3 voxel size; sequence duration of 3-4 minutes) was angled perpendicular to the long axis of the rectal cancer. Finally, an axial T1-weighted TSE sequence (656/10; echo-train length of 5; 5-mm slice thickness; 1-mm gap; 4 signals acquired; 256 x 256 matrix; 24-cm field of view; sequence duration of 4-5 minutes) was acquired. All sequences were obtained with no fat saturation. The total examination time for each patient ranged from 19 to 22 minutes.
Endorectal Sonographic Examinations
All endorectal sonographic examinations were performed with a real-time
scanner and a 10- and 7.5-MHz 360° rigid radial transducer (model
SSD-5500; Aloka) by one of two experienced abdominal radiologists. The
radiologists were provided with full knowledge of the results of endoscopic
examinations. Patients were asked to perform rectal cleansing 2 hours before
endorectal sonography with two rectal suppositories (bisacodyl) and then to
lie in the left lateral decubitus position. After 50-150 mL of degassed water
was instilled in the rectal lumen with an enema syringe
[15], the transducer was
inserted into the anus and was advanced into the rectum as deeply as possible.
As the transducer was slowly pulled back, serial images of the entire length
of the rectum and anal canal were obtained. The transducer used was equipped
with a mechanical rotating probe that provided 360° anorectal images. The
total examination time for each patient ranged from 7 to 20 minutes.
Histopathologic Examinations
After total mesorectal excision, each specimen was opened along the
anterior border proximal to the segment containing the tumor. Before the
specimens were fixed in formalin, a pathologist harvested the lymph nodes in
the mesorectum. All specimens, including those from transanal endoscopic
microsurgery, were fixed by total immersion in buffered formalin for 48 hours
and were sliced transversely at 3-mm intervals. The extent of local tumor
staging in each slice was assessed according to the tumor component of the TNM
system.
Image Analysis
Three experienced abdominal radiologists independently reviewed the MR and
sonographic images in random order and in a blinded manner. The interval
between reviews of the MR and of the sonographic images was at least 3 weeks.
All images were evaluated with a 2,000 x 2,000 PACS (Path-speed
workstation, GE Healthcare) with adjustment of the optimal window setting in
each case.
MRI allowed visualization and delineation of the layers of both the rectal wall and mesorectal fascia of all patients. The tumor had higher signal intensity on the T2-weighted images than did the muscle layer. The depth of cancer invasion on MRI (mT stage) was interpreted as follows: mT1, the tumor signal intensity was confined to the submucosal layer and had relatively low signal intensity compared with the high signal intensity of the surrounding submucosa; mT2, the tumor signal intensity extended to the muscle layer, leading to irregularity or thickening of the muscle layer, but without perirectal tissue invasion; mT3, the tumor signal intensity extended through the muscular layer into the perirectal tissue, or angiolymphatic tumor invasion (irregularly thickened strands) was present in the mesorectum; and mT4, the tumor signal intensity extended to an adjacent organ, mesorectal fascia, or bowel [16-18]. Rather than using the size criteria, we defined the criteria for lymph node metastasis as any size of indistinct border, irregular margin, or mixed signal intensity [19]. Observers recorded the number of metastatic lymph nodes at MRI of each patient. Using a confidence level scoring system, observers independently scored the MR images for muscularis propria invasion, perirectal tissue invasion, and lymph node involvement. The following confidence levels were used: 1, definitely absent; 2, probably absent; 3, possibly present; 4, probably present; and 5, definitely present.
Observers were able to detect hypoechoic rectal tumors on endorectal sonographic images. They evaluated depth of tumor invasion, which was adjusted for sonography as previously described [15, 20, 21]. The depth of cancer invasion at sonography (sT stage) was interpreted as follows: sT1, the middle echogenic layer (representing submucosa) was irregularly thinned by a hypoechoic mass; sT2, complete disruption of submucosa was present, often coupled with thickening of the outer hypoechoic layer (representing muscularis propria); sT3, the border between the muscularis propria and the outer echogenic layer (interpreted as perirectal tissue) was irregular or serrated; and sT4, the mass invaded an adjacent organ. Observers considered uniformly hypoechoic rounded or oval structures with discrete margins situated in the perirectal tissue and having a short-axis diameter larger than 3 mm to be metastatic lymph nodes, and their data were recorded [15, 19-21]. Using a five-point confidence scale, observers independently scored the endorectal sonographic images for muscularis propria invasion, perirectal tissue invasion, and lymph node involvement.
Statistical Analysis
The study coordinator correlated pathologic stage (pT staging and pN
staging) with the results of MRI and endorectal sonography analyzed by three
observers. For local invasion of cancer, we calculated the sensitivities and
specificities of endorectal sonography and MRI as follows: for muscularis
propria invasion, stage T2 or higher versus stage T1; for perirectal tissue
invasion, stage T3 or higher versus stage T2 or lower; and for invasion of
adjacent organs, stage T4 versus stage T3 or lower
[3]. In addition, to determine
the presence or absence of metastatic lymph nodes, we calculated and compared
the sensitivities and specificities of each technique. Using the McNemar test,
we assessed the differences between estimates for endorectal sonography and
MRI. A p value less than 0.05 was considered to indicate a
statistically significant difference.
The observer performances for predicting muscularis propria invasion, perirectal tissue invasion, and lymph node involvement were examined by analysis of receiver operating characteristic (ROC) curves. Composite ROC curves for the combined performance of all observers were obtained by application of the maximum-likelihood curve-fitting algorithm to the pooled data of the three observers for each imaging technique. The area under each ROC curve (Az) was used to indicate the overall performance of the imaging techniques.
|
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
Local Invasion
Histopathologic examinations revealed six T1 cancers, three T2 cancers, and
15 T3 cancers. No T4 tumors were found in any of the surgical specimens. For
muscularis propria invasion, the mean sensitivities of both MRI and endorectal
sonography for all observers were 100% (54 of 54), and the mean specificities
were 66.7% (12 of 18) and 61.1% (11 of 18), respectively
(Table 1). The differences in
mean sensitivities and specificities were not statistically significant
(p > 0.05 in each case). For three observers, six (10%) of 60 mT2
or mT3 cancers on MRI and seven (11.5%) of 61 sT2 or sT3 on endorectal
sonography were false-positive results (pT1); thus all of these tumors were
overstaged (Fig. 1A,
1B). The positive predictive
value of MRI was 90.0% (54 of 60) and that of sonography was 88.5% (54 of 61).
The negative predictive values of both techniques were 100% (12 of 12 and 11
of 11). For muscularis propria invasion, the accuracies of MRI and endorectal
sonography for all observers were 91.7% (66 of 72) and 90.3% (65 of 72),
respectively. In the results of ROC assessment of pooled data from the three
observers, the Az values of MRI (mean, 0.971; 95% CI,
0.901-0.995) and endorectal sonography (mean, 0.978; 95% CI, 0.912-0.997) for
muscularis propria invasion showed no difference in diagnostic accuracy
(p = 0.696).
|
For perirectal tissue invasion, the mean sensitivities of MRI and endorectal sonography for all observers were 91.1% (41 of 45) and 100% (45 of 45), respectively, and the difference between these sensitivities was not statistically significant (p = 0.125) (Table 1). The mean specificities of MRI and endorectal sonography for all observers were 92.6% (25 of 27) and 81.5% (22 of 27), respectively. Again, the difference was not statistically significant (p = 0.453). On MRI, for all observers, two (4.7%) of 43 mT3 cancers were overstaged (mT3-pT2) (Fig. 2A, 2B), and four (13.8%) of 29 mT2 or mT1 cancers were understaged (mT2-pT3) (Fig. 3A, 3B). On endorectal sonography, five (10%) of 50 sT3 lesions were overstaged. The positive predictive value of MRI was 95.3% (41 of 43) and that of sonography was 90.0% (45 of 50). The negative predictive value of MRI was 86.2% (25 of 29) and that of sonography was 100% (22 of 22). For perirectal tissue invasion, the accuracies of MRI and endorectal sonography for all observers were 91.7% and 93.1%, respectively. The observers did not consider any rectal lesion T4 cancer on either MR or endorectal sonographic images.
|
|
|
|
|
|
Lymph Node Involvement
Histopathologic examination revealed 225 lymph nodes from the rectal cancer
resection specimens of 21 patients who underwent total mesorectal excision. Of
these, 35 (15.6%) of the perirectal nodes were found to be metastatic.
Thirteen patients, including three who underwent transanal endoscopic
microsurgery, had N0 disease. Seven patients had N1 disease (up to three
metastatic nodes), and four had N2 disease (four or more metastatic
nodes).
For lymph node involvement, the mean sensitivities of MRI and endorectal sonography for all observers were 63.6% (21 of 33) and 57.6% (19 of 33), and the mean specificities were 92.3% (36 of 39) and 82.1% (32 of 39), respectively (Table 1). The differences in mean sensitivities and specificities were not statistically significant (p > 0.05 in each case). The positive predictive value of MRI was 87.5% (21 of 24), and the negative predictive value was 75% (36 of 48). The positive predictive value of endorectal sonography was 73.1% (19 of 26), and the negative predictive value was 69.6% (32 of 46). The accuracies of MRI and endorectal sonography were 79.2% (57 of 72) and 70.8% (51 of 72), respectively. The ROC curves for MRI (mean, 0.776; 95% CI, 0.663-0.866) and endorectal sonography (mean, 0.721; 95% CI, 0.603-0.821) for lymph node involvement showed no difference in diagnostic accuracy (p = 0.43).
The kappa values for the three observers of lymph node involvement showed moderate agreement for MRI (Table 2). For endorectal sonography, the kappa values for the three observers showed good or excellent agreement.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
In patients with rectal cancer, CT is an important imaging technique for preoperative staging of colorectal cancer because it allows evaluation for distant metastasis. CT, however, is limited in depicting differentiation of the layers of the rectal wall, showing the mesorectal fascia, and depicting cancer invasion into adjacent organs. MDCT may improve the diagnostic value of CT, because it allows assessment of local disease with improved visualization of the mesorectal fascia [26]. The increased speed and spatial resolution of MDCT may allow multiplanar reconstruction without stepping artifacts. Multiplanar reconstruction in the coronal and sagittal views helps in evaluation of adjacent organs for invasion of rectal cancer [27].
Some investigations have reported good results with high-spatial-resolution MRI performed at 1.5 T with phased-array coils [16, 28, 29]. Brown et al. [16] reported that the accuracy of staging in rectal cancer at 1.5 T was 100% with complete agreement between two observers. However, in their study, the total examination time required for good image resolution was too long, ranging from 45 to 65 minutes. This problem can be overcome with a 3-T MRI system. Because the increase in signal-to-noise ratio at 3-T can decrease the number of signals averaged, the total MRI examination time for scanning at 3 T is quite a bit shorter than that at 1.5 T [11-14]. In our study, the MRI examination time for each patient ranged from 19 to 22 minutes. In addition, the higher signal-to-noise ratio of 3-T MRI can improve spatial resolution and allow acquisition of thinner sections. Because signal-to-noise ratio is not solely determined by magnetic field strength, optimization of the imaging parameters and improvement of phased-array radiofrequency coils are important for maximal signal-to-noise ratios in body imaging [12].
According to a meta-analysis of 90 articles [3], the specificity of sonography (95% CI, 80-90%) for muscularis propria invasion is significantly higher than that of MRI (95% CI, 52-82%). In our study, however, the differences in specificities were not statistically significant. This discordance is most likely a result of the small numbers (9 of 24, 37.5%) of T1 and T2 cancers in our study. For perirectal tissue invasion in the meta-analysis, the sensitivity of sonography (95% CI, 88-92%) was significantly higher than that of MRI (95% CI, 74-87%), and the specificities were comparable. In ROC assessment of diagnostic accuracy for perirectal tissue invasion in the meta-analysis, sonography was found to be better than MRI. Our study also showed endorectal sonography to be more accurate than MRI in assessment of perirectal tissue invasion.
In patients with rectal cancer, identification of metastatic lymph nodes with endorectal sonography, CT, and MRI is still a challenge for radiologists [3]. For lymph node involvement, estimates of sonography and MRI were comparable in the meta-analysis and in our study. Sonography can be used to assess only perirectal lymph nodes, whereas MRI can also be used to evaluate periiliac lymph nodes.
The diagnostic criteria for nodal metastasis continue to be debated. Some authors regard any visible node in the perirectal tissue as a positive finding, whereas others use size criteria with a range of 3-10 mm [30, 31]. Brown et al. [19] found that an irregular border or mixed signal intensity of lymph nodes at MRI can improve prediction of the presence of nodal metastasis regardless of the size of the lymph nodes. Despite detection of lymph nodes as small as 2-3 mm on high-spatial-resolution images, identification of nodal metastasis is uncertain. Although the spatial resolution at 3-T MRI was increased in our study, the sensitivity in detection of nodal metastasis was as low as 63.6% and could not be improved [8, 32, 33]. The morphologic criteria used in our study made it difficult to differentiate reactive and metastatic nodes. To improve the sensitivity of MRI in lymph node detection, techniques such as use of lymph node-specific MRI contrast agents show promising results in staging of nodal metastasis [34].
Although it was found better than MRI in our study, endorectal sonography has drawbacks. First, sonographic examinations depend on the experience of the operators of the equipment. Second, it is exceedingly difficult, almost impossible, to evaluate stenotic tumors and tumors situated 10 cm from the anal verge. In addition, selection of patients can lead to biased results. In our study, overstaging was the most frequent cause of inaccurate staging with endorectal sonography, a finding consistent with those in the literature [35, 36]. The presence of inflammatory reactions and desmoplastic changes around tumors can cause overstaging because these changes cannot be differentiated from cancer infiltration [36, 37]. In our study, endorectal sonography with water instillation into the rectal lumen was used to obtain an optimal sonic window [15]. This method is thought to be a reliable method of reducing overstaging of rectal cancers because it decreases artifacts originating from the tumor itself or from feces. In addition, by using the water instillation method, we were able to easily perform endorectal sonography in a distended rectal lumen and could advance the transducer into the upper rectum.
Our study had several limitations. The first was the small population, particularly the small number of T1 and T2 lesions (9 of 24, 37.5%). Second, patients with T4 and advanced T3 cancers (cancer infiltration near mesorectal fascia) were not included. For these patients, preoperative radiation therapy is considered the standard of treatment at many institutions. Thus, these patients underwent chemoradiation therapy before surgery in our hospital. It is difficult to correlate findings on imaging before radiation therapy with the findings in surgical specimens after radiation owing to shrinkage of the rectal cancer, postirradiation edema of the rectal wall, and interval regression of perirectal lymph nodes [38, 39]. Finally, despite water filling into the rectum before MRI, some patients did not have a fully distended rectum at MRI. In these patients, it was difficult to evaluate perirectal tissue invasion [40].
In conclusion, in our preliminary study, despite use of a newly developed 3-T system, MRI was less accurate than endorectal sonography in the detection of perirectal tissue invasion.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  T. Leibold, J. Shia, L. Ruo, B. D. Minsky, T. Akhurst, M. J. Gollub, M. S. Ginsberg, S. Larson, E. Riedel, W. D. Wong, et al. Prognostic Implications of the Distribution of Lymph Node Metastases in Rectal Cancer After Neoadjuvant Chemoradiotherapy J. Clin. Oncol., May 1, 2008; 26(13): 2106 - 2111. [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]
|
|
|
|
 |
|
 S. I. Sher Invited Commentary RadioGraphics, September 1, 2007; 27(5): 1462 - 1464. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
) and
endorectal sonography (
) calculated with pooled data reviewed by three
observers. Mean area under ROC curve (Az) indicates diagnostic
accuracy for perirectal tissue invasion at endorectal sonography
(Az = 0.982 ± 0.006) and at 3-T MRI (Az
= 0.952 ± 0.010). Difference in mean Az was
statistically significant (p = 0.028).