Colorectal carcinoma is one of the most common malignant neoplasms worldwide, and the prognosis is closely correlated with the depth of invasion and the presence of lymph node involvement [
Magnetic resonance (MR) imaging also has been used to stage rectal cancer, but previous reports suggest that although conventional MR imaging provides high soft-tissue contrast, it has substantial limitations in regard to T- (primary tumor) staging and N- (regional lymph nodes) staging of rectal cancer because of its limited spatial resolution [
The materials consisted of 92 surgical specimens containing 96 colorectal tumors that had been obtained from 92 consecutive colorectal cancer patients at our institution and that had been histopathologically confirmed to be adenocarcinoma. Sixty-two of the patients were male and 30 were female, and their ages at the time of surgery ranged from 42 to 87 years (mean age: 65 years ± 9 [standard deviation]). The cancers were located in the rectum (
High-resolution MR imaging was performed by using a 1.5 T system with a 25 mT/m maximum gradient capability (Magnetom Vision; Siemens, Erlangen, Germany) and a 4 cm diameter loop coil. All specimens were imaged in vitro after fixation in formalin. Conventional single-section sagittal, coronal, and axial scout images of the colorectal specimen were initially obtained.
High-resolution T1-weighted spin-echo MR images were obtained with a 500/20 (repetition time msec/echo time msec) sequence and with eight signals acquired. High-resolution T2-weighted spin-echo MR images were obtained with a 2000/70 sequence and with four signals acquired. All images were obtained with a 50
The MR images of each lesion were interpreted by two independent radiologists (I.Y., N.Y.), who were blinded to the histopathologic findings. The histopathologic findings were used as the reference standard for analysis of the MR imaging findings. When the radiologists did not fully agree on the findings, the final determination was made by consensus.
MR imaging criteria used to determine the depth of invasion in colorectal carcinoma.
Depth of invasion | MR imaging criteria |
---|---|
Mucosa (Tis) | Thickening in the mucosal layer |
Submucosa (T1) | Mass in the submucosal layer |
Muscularis propria (T2) | Mass extending into the muscle layer |
Abnormal signal intensity in the thickened muscle layer | |
Subserosa/serosa or adventitia (T3 and T4) | Mass extending through the muscle layer into the subserosa or adventitia |
Note. Letters in parentheses indicate the corresponding tumor stage according to the International Union against Cancer Tumor-Node-Metastasis classification [
Concerning matching the imaging findings with the pathology findings, we evaluated the deepest invasion area in each tumor lesion on MR images and pathologic sections separately, so as to determine the T-stages of imaging and pathology independently. Thus, the T-stage of MR imaging was matched with the T-stage of pathology in each tumor lesion of the specimens.
After MR imaging the surgical specimen was sectioned longitudinally so that it corresponded to the orientation of the MR images, and the sectioned specimens were embedded in paraffin and cut into 6
The sensitivity, specificity, and accuracy of high-resolution MR imaging as a method of assessing the depth of carcinoma invasion and lymph node metastasis were determined by comparison with the histopathologic findings. Depth of invasion according to the MR imaging findings and the histopathologic findings was compared by using the Spearman correlation coefficient. The nodal size of metastatic and nonmetastatic lymph nodes was compared by using Student's
High-resolution T2-weighted MR images depicted the mucosa as low signal intensity, and the muscularis mucosae, the deepest layer of the mucosa, as a separate layer that had a lower signal intensity than other parts of the mucosa (Figure
Images of the normal colorectal wall. (a) High-resolution T2-weighted MR image (2000/70) clearly depicts the normal colorectal wall as consisting of seven layers, which correspond well with the histopathologic layers.
Thus, the high-resolution T2-weighted MR images clearly depicted the normal colorectal wall as consisting of the following seven layers that correlated well with the layers of the colorectal wall histopathologically: mucosa (low signal intensity), muscularis mucosae (low signal intensity), submucosa (high signal intensity), inner circular muscle layer (low signal intensity), intermuscular connective tissue (high signal intensity), outer longitudinal muscle layer (low signal intensity), and subserosa/serosa or adventitia (high signal intensity).
The high-resolution T1-weighted MR images depicted the mucosa, muscularis mucosae, submucosa, muscularis propria, and subserosa/serosa or adventitia in the colorectal wall as having similar low signal intensity, and the fat tissue in the submucosa and subserosa or adventitia as high signal intensity (Figure
On the high-resolution T2-weighted MR images, the colorectal wall appeared as seven layers in 44 (48%) of the 92 specimens. In 46 (50%) specimens, however, the colorectal wall appeared as six layers because the muscularis mucosae was not separated from the mucosa. In the remaining two (2%) specimens, only four layers were observed because the muscularis mucosae was not separated from the mucosa and the muscularis propria appeared as a single low-signal-intensity zone.
At histopathologic examination, the 96 colorectal carcinomas in this series consisted of 19 carcinomas confined to the mucosa, 15 that had invaded the submucosa, 15 that had infiltrated the muscularis propria, and 47 that had extended into the subserosa/serosa or adventitia (Table
Comparison of high-resolution MR imaging and histopathologic findings for evaluating the depth of invasion in colorectal carcinoma.
Histopathologic findings | ||||
MR imaging findings | Mucosa( | Submucosa ( | Muscularis propria ( | Subserosa/serosa or adventitia ( |
Mucosa | 17 | 0 | 0 | 0 |
Submucosa | 2 | 14 | 0 | 0 |
Muscularis propria | 0 | 1 | 13 | 1 |
Subserosa/serosa or adventitia | 0 | 0 | 2 | 46 |
Note. Numbers are numbers of lesions among 96 carcinomas in 92 patients.
Colorectal carcinoma confined within the mucosa. (a) High-resolution T2-weighted MR image (2000/70) shows an irregular thickening (arrows) in the mucosa, and the submucosa of high signal intensity appears to be intact. (b) Corresponding histopathologic section shows carcinoma confined within the mucosa (arrows) as well as intact submucosa. (Hematoxylin-eosin stain; original magnification:
Colorectal carcinoma invading the submucosa. (a) High-resolution T2-weighted MR image (2000/70) shows that an irregularly-shaped tumor (arrows) contrasts with the high-signal-intensity submucosa. (b) Corresponding histopathologic section shows carcinoma invading the submucosa (arrows). (Hematoxylin-eosin stain; original magnification:
Colorectal carcinoma involving the muscularis propria. (a) High-resolution T2-weighted MR image (2000/70) shows that an irregularly-shaped tumor partially replaces the muscularis propria layer (arrow), but that it does not penetrate through the muscularis propria layer. There is a deep ulceration in the central part of the tumor. (b) Corresponding histopathologic section shows carcinoma involving the muscularis propria (arrow) which manifests a deep ulceration in the central part. (Hematoxylin-eosin stain; original magnification:
The depth of carcinoma invasion of the colorectal wall was clearly demonstrated by high-resolution T2-weighted MR imaging. On the high-resolution T2-weighted MR images carcinomas confined to the mucosa were visualized as a discrete low-signal-intensity thickening in the mucosal layer (Figure
As shown in Table
Table
Diagnostic accuracy of high-resolution MR imaging for evaluating the depth of invasion in colorectal carcinoma.
Depth of invasion | Sensitivity | Specificity | Accuracy |
---|---|---|---|
Mucosa | 96/96 (100) | 0/0 (NA) | 96/96 (100) |
Submucosa | 77/77 (100) | 17/19 (89) | 94/96 (98) |
Muscularis propria | 62/62 (100) | 33/34 (97) | 95/96 (99) |
Subserosa/serosa or adventitia | 46/47 (98) | 47/49 (96) | 93/96 (97) |
Note. Data represent the diagnostic accuracy for invasion of each layer of the colorectal wall in 96 carcinomas. Values in parentheses are percentages. NA: not applicable.
A Spearman coefficient (
Since high-resolution MR imaging depicted 82 lymph nodes in the 92 colorectal specimens that were examined histopathologically, we compared the high-resolution MR imaging findings and histopathologic findings in the 82 lymph nodes. The 82 lymph nodes confirmed by pathology were in 33 (36%) of the 92 surgical specimens. The number of lymph nodes in these specimens ranged from one to eight, and the median number of lymph nodes was two (the mean number: 2.5). Other lymph nodes also were separately harvested at surgery from the 92 patients, but these lymph nodes were not analyzed because they were not imaged with high-resolution MR imaging. Histopathologic examination revealed metastasis in 32 (39%) of the 82 lymph nodes and no metastasis in the other 50 (61%) lymph nodes.
The evaluation of nodal size showed that the metastatic lymph nodes ranged from 2.6 mm to 12.6 mm in size and that the benign lymph nodes ranged from 1.5 mm to 6.9 mm in size (Figure
The size, signal intensity, and border contour of lymph nodes in colorectal carcinomas on high-resolution MR images. (a) Box plot shows the diameter (measured in millimeters) of benign and metastatic lymph nodes on high-resolution MR images. The horizontal line within the box is the median value (50th percentile), the boundaries of the box represent 25th and 75th percentiles, and whiskers show 10th and 90th percentiles. Values above the 90th and below the 10th percentiles are plotted as data points (circles). (b) Bar chart shows the number of benign (white bars) and metastatic (black bars) lymph nodes on high-resolution MR images, according to signal intensity on T2-weighted MR images. (c) Bar chart shows the number of benign (white bars) and metastatic (black bars) lymph nodes on high-resolution MR images, according to border contour.
The evaluation of signal intensity revealed metastasis in 23 (92%) of the 25 lymph nodes showing mixed signal intensity on high-resolution T2-weighted images and no metastasis in the other two (8%) lymph nodes (Figures
Metastatic lymph nodes showing mixed signal intensity on high-resolution T2-weighted images and irregular border contour. (a) High-resolution T2-weighted MR image (2000/70) shows lymph nodes (arrows) having mixed signal intensity and irregular border contour in the subserosal fat. (b) High-resolution T1-weighted MR image (500/20) shows lymph nodes (arrows) having irregular border contour. (c) Corresponding histopathologic section shows metastatic lymph nodes (arrows) in the subserosa. (Hematoxylin-eosin stain; original magnification:
Metastatic lymph nodes showing low signal intensity on high-resolution T2-weighted images and irregular border contour. (a) High-resolution T2-weighted MR image (2000/70) shows lymph nodes (arrows) having low signal intensity and irregular border contour in the subserosal fat. (b) High-resolution T1-weighted MR image (500/20) shows lymph nodes (arrows) having irregular border contour. (c) Corresponding histopathologic section shows metastatic lymph nodes (arrows) in the subserosa. (Hematoxylin-eosin stain; original magnification:
Benign lymph nodes showing high signal intensity on high-resolution T2-weighted images and smooth border contour. (a) High-resolution T2-weighted MR image (2000/70) shows lymph nodes (arrow) having high signal intensity and smooth border contour in the subserosal fat. The chemical shift artifact (arrowheads) is noted. (b) High-resolution T1-weighted MR image (500/20) shows lymph nodes (arrow) having smooth border contour. The chemical shift artifact (arrowheads) is noted. (c) Corresponding histopathologic section shows benign lymph nodes (arrow) in the subserosa. (Hematoxylin-eosin stain; original magnification:
The evaluation of border contour showed that 29 (91%) of the 32 lymph nodes with an irregular, ill-defined border contour were metastatic, and the other three (9%) were metastasis-free (Figures
Table
Diagnostic accuracy of high-resolution MR imaging for evaluating lymph node metastasis in colorectal carcinoma.
Criterion for lymph node metastasis | Sensitivity | Specificity | Accuracy |
Nodal size | |||
≥8 mm | 7/32 (22) | 50/50 (100) | 57/82 (70) |
≥7 mm | 11/32 (34) | 50/50 (100) | 61/82 (74) |
≥6 mm | 21/32 (66) | 49/50 (98) | 70/82 (85) |
≥5 mm | 26/32 (81) | 46/50 (92) | 72/82 (88) |
≥4 mm | 29/32 (91) | 44/50 (88) | 73/82 (89) |
≥3 mm | 31/32 (97) | 29/50 (58) | 60/82 (73) |
Signal intensity (SI) | |||
Mixed | 23/32 (72) | 48/50 (96) | 71/82 (87) |
Mixed or iso/low | 32/32 (100) | 44/50 (88) | 76/82 (93) |
Border contour | |||
Irregular | 29/32 (91) | 47/50 (94) | 76/82 (93) |
Mixed or iso/low SI and irregular border | |||
29/32 (91) | 50/50 (100) | 79/82 (96) | |
Nodal size ≥4 mm and (mixed or iso/low SI or irregular border) | |||
29/32 (91) | 48/50 (96) | 77/82 (94) |
Note. Values in parentheses are percentages.
Both univariate and multivariate logistic regression analyses revealed nodal size, signal intensity, and border contour as significant parameters for predicting lymph node metastasis. The multivariate logistic regression analysis demonstrated the model which could best predict lymph node metastasis from nodal size, signal intensity, and border contour, with the following equation for the logit:
Our findings demonstrated that high-resolution T2-weighted MR images clearly depicted the normal colorectal wall as consisting of seven layers that corresponded well with the actual layers of the colorectal wall observed histopathologically. High-resolution MR imaging provides much higher soft-tissue contrast than CT or US, and there are none of the artifactual interface echoes in the colorectal wall that occur with US. Previous studies have described the colorectal wall as consisting of three to six layers on T2-weighted MR images [
Our findings showed that high-resolution MR imaging was able to correctly depict the depth of invasion of the colorectal wall in 90 (94%) of the 96 colorectal carcinomas studied. Although the assessments of depth of invasion by MR imaging resulted in overestimation in five (5%) of the other carcinomas and underestimation in other one (1%), the ranges of its sensitivity, specificity, and accuracy as a method of assessing depth of invasion of the colorectal wall were 98%–100%, 89%–97%, and 97%–100%, respectively. Thus, high-resolution MR imaging was found to be a highly accurate method for evaluating depth of invasion by colorectal carcinoma. Its high accuracy appears to be attributable to the combination of high soft-tissue contrast and high spatial resolution that high-resolution MR imaging provides [
Our findings also demonstrated a high degree of accuracy of high-resolution MR imaging for evaluating lymph node metastasis by colorectal carcinoma. Nodal signal intensity on high-resolution T2-weighted images (93%) and nodal border contour (93%) provided greater accuracy than nodal size (89%). MR imaging yielded the greatest accuracy (94%–96%) for evaluating lymph node metastasis when nodal size, signal intensity and border contour were combined.
Previous reports indicate that lymph node evaluation in rectal cancer is challenging for every imaging technique, because lymph node size alone is not a reliable diagnostic criterion for metastatic involvement [
A limitation of our study is that the specimens were imaged after fixation in formalin. However, since previous reports have shown no substantial effect of formalin fixation, on the signal intensity and soft-tissue contrast of T2-weighted images in the colorectal wall, gastric wall, and esophageal wall [
Another limitation of our study is that we analyzed only the pericolorectal lymph nodes adjacent to the primary tumor in the specimens, and thus more proximal lymph nodes were not analyzed, even when separately harvested at surgery. However, this procedure enabled strict node-by-node correlations to be made with the histopathologic findings, and in our study it was possible to strictly determine whether individual lymph nodes on high-resolution MR images contained metastases. Studies using a patient-by-patient analysis alone might be insufficient to clarify imaging findings of individual lymph nodes containing metastases [
Finally, since our results were obtained by imaging surgical specimens using a dedicated coil and long acquisition times, the results cannot be directly extrapolated to clinical practice. There are many technical issues associated with performing the high-resolution technique in vivo, including bowel peristalsis, patient motion, bowel collapse, residual fecal material and gas, and distance from the coil to the lesion. However, high-resolution MR imaging in vivo may become possible by using an endoluminal coil technique or phased-array coil technique and with the development of faster MR imaging techniques. Higher field strength (3.0 T) also may enable high-resolution MR imaging in reduced acquisition times.
In conclusion, our study has demonstrated that high-resolution MR imaging clearly depicts the internal architecture of the colorectal wall in surgical specimens and is a highly accurate diagnostic method for evaluating depth of invasion and lymph node metastasis by colorectal carcinoma. Thus, high-resolution MR imaging may enable accurate preoperative local tumor staging and lymph node assessment of colorectal carcinomas.