| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
1
Department of Imaging, Hammersmith Hospital, Imperial College School of
Medicine, Du Cane Rd., London W12 0HS, United Kingdom.
2
Department of Gynecological Oncology, Hammersmith Hospital, Imperial College
School of Medicine, London W12 0HS, United Kingdom.
3
Department of Histopathology, Hammersmith Hospital, Imperial College School of
Medicine, London W12 0HS, United Kingdom.
Received November 27, 2000;
accepted after revision February 2, 2001.
Presented at the annual meeting of International Society for Magnetic
Resonance in Medicine, Denver, April 2000.
Abstract
 Top Abstract IntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
MATERIALS AND METHODS. Eighteen patients with gynecologic cancers were studied by all three imaging methods before surgery. The images were initially reviewed with routine diagnostic conditions and then, subsequently, by two observers who were unaware of the clinical and histopathologic findings of the patients. The nodal sites were split into upper (aortic to common iliac bifurcations) and lower (common iliac bifurcations to inguinal ligament) iliac chains. All observers' results were statistically analyzed with specificity, sensitivity, positive and negative predictive values, Fisher's exact test (individual observers) or chi-square test (combined observers), and Cohen's kappa test.
RESULTS. Eight of 18 patients had lymph node metastases at histology. Findings of all three modalities agreed in full in only one patient. CT correctly revealed 10 node-negative patients, whereas MR imaging was correct in eight of these patients. 18F-FDG PET correctly depicted one patient with lymph nodes negative for tumor. CT was the most specific imaging modality (97.0%), with MR imaging and PET rendering values of 90.7% and 77.3%, respectively, but sensitivity of all modalities was low (CT, 48.1%; MR imaging, 53.7%; PET, 24.5%). Observer agreement for each modality was good; kappa values among all observers were 0.88 for CT, 0.85 for MR imaging, and 0.72 for PET.
CONCLUSION. CT is the most specific modality for detecting lymph nodes positive for tumor in gynecologic cancers, whereas MR imaging is the most sensitive. The poor results of PET in the pelvis are attributed to urinary 18F-FDG in the ureters or bladder, which may mask or imitate lymph node metastases.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CT
Patients had bowel preparation with oral meglumine diatrizoate
(Gastrografin; Schering Health Care, West Sussex, United Kingdom) (500 mL of a
2.5% solution) 2 hr before scanning. Images were obtained with Somatom Plus
(eight patients) or Somatom Plus 4 (10 patients) scanners (Siemens, Erlangen,
Germany). The patients were scanned from the symphysis pubis to the apex of
the lungs, and all scans were contrast-enhanced with 100 mL of iopomide
(Ultravist 300; Schering Health Care)
[9]. In the Somatom Plus
scanner, 10-mm contiguous slices were obtained for the pelvis and abdomen, and
10-mm reconstructed helical CT scans were obtained for the chest
[10,11].
Using the Plus 4 helical scanner, we obtained slices with 10-mm thickness
image reconstruction in two acquisitions, pelvis to diaphragm and then chest.
This procedure ensured that the liver was imaged in the portal phase.
The initial images were reported by 11 radiologists (initial observer).
MR Imaging
Imaging was performed on a 0.5-T super-conducting magnet (Asset; Marconi
Medical Systems, Highlands Heights, OH) with a pelvic phased array imaging
coil. No prior patient preparation was used. At the start of scanning, each
patient received 20 mg of intramuscular hyoscine-N-butylbromide (Buscopan;
Boehringer-Ingelheim, Berks, United Kingdom) to reduce bowel motion artifact.
The 35-cm field of view covered from the symphysis pubis to the bifurcation of
the aorta. Coronal T1-weighted spin-echo (TR/TE, 638/20), short tau inversion
recovery fat-suppression sequences (2000/30; inversion time, 107 msec),
sagittal T2-weighted fast spin-echo (TR/effective TE, 4000/88), and transverse
T1-weighted spin-echo (709/20) images with a slice thickness of 8 mm were
obtained.
Initially, the images were reviewed with routine clinical reporting conditions by one radiologist (initial observer) using an Alpha workstation (Digital Electronic Corporation, Houston, TX).
PET
PET whole-body images were obtained with an ECAT ART scanner (Siemens-CTI,
Knoxville, TN) with a 16.2-cm axial field of view (reconstructed slice
thickness, 10 mm; spatial resolution, 6 mm). The first five patients had no
preparation, but the subsequent 13 fasted for at least 12 hr before the scan
and had their blood glucose measured. None were excluded on the basis of
hyperglycemia. The final 13 patients were well hydrated and given 20 mg of
furosemide (Lasix; Borg Medicare, Herts, United Kingdom) intramuscularly to
aid renal washout. Patients were injected with 88-222 MBq of
18F-FDG (mean, 165 MBq), and imaging was performed 37-102 min after
injection (mean, 55.6 ± 25.3 min). Patients emptied their bladders
immediately before scanning. Studies were performed with a whole-body
technique, mid thighs to ears, with a scan time of 45 min. Supine patients
were scanned in 6-7 bed positions with 7 min in each position. Patients were
scanned without initial transmission scans, so final images were not
attenuation-corrected [12].
All images were displayed on a workstation monitor as three-dimensional
black-and-white images.
Two independent nuclear medicine physicians (initial observer) reviewed the studies initially with routine reporting conditions.
Image Analysis
Subsequently, independent observers, unaware of patient clinical and
histopathologic details, assessed images formally for the purposes of this
study. For each modality, an observer (first observer) with specialist
experience in that particular modality and an observer (second observer) with
a more general interest were used. Images were reviewed from either a soft or
a hard copy, according to observer preference. Observers used prompt sheets
specifying the nodal sites to ensure uniform responses. The nodal regions
scored were divided into left and right and upper and lower iliac chain
areas.
The upper iliac lymph node chain extended from aortic to iliac bifurcations; the lower iliac node chains extended from the common iliac bifurcations to the inguinal ligaments. These divisions were chosen because they could be easily identified on PET scans and simplified the imaging analysis.
The patients imaged were not consecutive. They were referred for PET scanning on the basis of large stage Ib cervical cancer, small tumors with parametrial extension, or recurrent pelvic disease. Imaging with PET was also dependent on staffing and radiopharmaceutical availability, because this was a new modality in the main imaging department.
Nodal positivity was determined by size on MR imaging and CT; nodes greater than 1-cm long axis were noted as positive for tumor. Findings of focal 18F-FDG uptake along the iliac chains were noted as positive on PET.
Histology
At pelvic lymph node dissection, all tissues were labeled for site and side
and documented by the senior operating room technician or assisting clinician.
The surgical specimens were fixed in formalin.
The lymph nodes were macroscopically described for size, form, and nature and then embedded in paraffin wax. They were bisected before embedding in paraffin, routinely processed, and stained with H and E before microscopic examination. Any metastases identified were typed to ensure that they were from the primary tumor. Histology was obtained for every lymph node dissected, and all tumors were staged according to the international primary TNM classification [13]. The histologic slides were retrospectively reviewed in all patients. The maximal lengths of all available lymph nodes with positive and negative findings were recorded with a centimeter scale.
Statistical Analysis
Comparison with histology was made for right and left and upper and lower
iliac regions for each imaging modality and for each observer. Thus, the
number of true-positives, true-negatives, false-positives, and false-negatives
for every patient in each of four nodal sites was tabulated.
The sensitivity, specificity, and positive and negative predictive values were calculated with standard statistical formula. Cohen's kappa tests [14] were applied to test observer agreement in each modality. Each observer's results for both iliac sites and combined sites were subjected to a Fisher's exact test, a 2 x 2 table with positive/negative findings on histology versus each reviewer's observations. Chi-square tests, a 2 x 2 table with positive and negative findings on histology versus observers, were used for the combined results for all observers of each modality to test the hypothesis that the observers had not arrived at their agreement with histology by chance. A p value of 0.05 was chosen to indicate a 95% probability that the results were achieved by expert observation of the imaging modality.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
The number of true-positives, true-negatives, false-positives, and false-negatives for every patient in each of four nodal sites was tabulated. Table 1 shows the totals of these figures for each modality, as upper, lower, and combined results.
|
Tables 2,3,4 give the sensitivity and specificity results for each radiologist and the combined results for all radiologists for each modality. Table 2 is for the upper, and Table 3 is for the lower iliac region. Sensitivity and specificity for upper and lower iliac sites were similar for each modality. Overall, CT achieved the highest specificity, 97.0%, whereas specificity with MR imaging was 90.7% and with PET, 77.3%. Although the sensitivity was poor on CT (48.1%) and MR imaging (53.7%), it was low for PET (24.5%) (Table 4). A patient with findings of true-positive nodes is illustrated in Figure 1A,1B,1C. Examples of false-positive nodes are shown in Figures 2A,2B and 3.
|
|
|
|
|
|
|
|
|
Positive and negative predictive values are given in Table 5. The results for each modality were similar for upper and lower iliac sites. For both iliac sites combined, CT had the highest positive predictive value (0.83) and a negative predictive value equal to MR imaging (0.85), whereas PET had the lowest positive predictive value (0.26), because of the high number of false-positive findings.
|
Observer Agreement
The results of Cohen's kappa tests of agreement among all radiologists for
each modality are given in Table
6.
|
For CT, the agreement among all observers for upper and lower iliac node
sites combined was good, with kappa values for observer combinations varying
between 0.85 and 0.94. CT had the highest agreement among all observers
(
= 0.88). For MR imaging, kappa values were also high (0.81-0.87).
Agreements among observers for the nodal sites combined in PET were good,
albeit slightly lower ( = 0.64-0.79).
The Fisher's exact test indicated that it was unlikely that any CT observers would have reached their agreement with histology by chance, because all observers had probability scores of less than 0.05. With MR imaging, the Fisher's exact test also indicated that observers were unlikely to have arrived at their results by chance (p < 0.05). For PET, the results of Fisher's exact test for upper and lower iliac sites separately and for the iliac sites combined was more than 0.05, indicating a greater than 5% probability that all observers could have arrived at their results by chance.
Agreement Among Modalities
Findings on all three modalities agreed on all sites in only one
node-negative patient. CT correctly reported all 10 node-negative patients,
whereas findings on MR imaging agreed in eight of these. 18F-FDG
PET correctly showed one patient with a true-negative lymph node status.
Histologic Analysis
Five hundred and four lymph nodes were resected and documented.
Macroscopic-length measurements in 33 of 34 nodes containing metastatic
deposits were obtained (Fig.
4). The mean length of positive nodes was 1.1 ± 0.6 cm
(median, 0.9 cm; range, 0.2-2.3 cm). Of the nodes recorded as positive for
tumor, 54.5% were less than 1.0 cm in length (45.5% 
1.0 cm in length).
Four hundred seventy nodes were negative for tumor, and it was possible to
obtain length measurements in 314 of these. The mean length of negative nodes
was 0.7 ± 0.5 cm (median, 0.6-cm; range, 0.1-2.5 cm). Of the nodes
negative for tumor, 22.6% were equal to or greater than 1.0 cm in length
(77.4% > 1.0 cm in length). The difference in the length of positive and
negative lymph nodes recorded on macroscopy was significant (p =
0.0).
|
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
With the advent of newer imaging modalities such as MR imaging and PET, it was hoped that other criteria such as the signal intensity [17] of the node or its metabolic status [5,6,7] might prove a more sensitive indicator of disease. However, preliminary results have been disappointing, and morphologic criteria alone are used. Further refinements include contrast enhancement [18, 19] and studies of oxygenation of the node [20], but even in a research setting, these parameters remain unreliable. This study, therefore, used size criteria alone to detect lymph nodes positive for tumor with respect to CT and MR imaging. As in others, this study also used a 1-cm-long axis cutoff [5, 16]. A lower figure would reduce the specificity, particularly in the pelvis where adjacent bowel creates difficulties with partial volume effects.
The positive predictive value of CT was higher than that of the other two
modalities. CT was, thus, of intermediate sensitivity, but highly specific. On
MR imaging, observers obtained equal or slightly higher sensitivity ratings
than on CT, because contrast between tissues was better on T2-weighted MR
sequences. However, specificity was slightly lower. The positive predictive
values for MR imaging were, therefore, low, because both observers had a
number of false-positive findings. Low sensitivity and specificity results
were obtained overall for PET imaging. These results were disappointing,
because PET has become more widely used for staging of head and neck cancers,
in lung cancer, and in malignant melanoma
[5,6,7,8,21].
This study emphasized the many problems facing 18F-FDG PET in the
pelvis [22]. Residues of
18F-FDG can sometimes be seen in the ureters and in the bladder,
masking or imitating metastases. These technical problems must be addressed to
improve the specificity of the technique
[23].
|
The requirement for histologic correlation after node dissection introduced bias into the study. Patients with unequivocally enlarged nodes would not have undergone surgical lymph node dissection but would have been selected for radiation therapy and would, therefore, have been excluded from analysis. Also, patients with small primary tumors and nodes not enlarged on initial investigation with cross-sectional imaging may not have been referred routinely for PET scanning. The population selected will, thus, have been patients destined for surgical lymph node dissection with a likelihood of small lymph node metastases.
Observer bias was minimized by using a uniform prompt sheet and by ensuring the first and second observers were unaware of the clinical and histopathologic details. CT is a well-established imaging modality, and as expected, the overall agreement among radiologists reviewing CT was high. In view of the good agreement among multiple observers, it can be concluded that the shortcomings of this modality are due to the choice of criteria for detection of the lymph nodes positive for tumor and technical factors rather than to observers' abilities. MR imaging is also routinely used for pelvic imaging, and, as with CT, the agreement among radiologists reviewing this modality was good. With PET, observers agreed on 11 patients for both the upper and lower iliac regions. This was good agreement in the modality, but only four agreed observations were correct compared with histology. There were no true-positive reports on which all three observers agreed. The use of PET in the pelvis is a relatively new phenomenon [24], and the poor results may be attributed partly to lack of observer experience.
CT has the advantages of better total-body coverage and short scanning time. Radiation dose is not an issue in this group of patients. The limitations of MR imaging are long scanning times and poor total-body coverage, but it offers superior definition of the primary disease. PET performs poorly in the pelvis because of inherent problems of imaging the pelvis with 18F-FDG. These problems currently are being addressed with further improvements in patient preparation. A combination of morphologic and functional techniques or multitracer PET studies may improve sensitivity further.
Cross-sectional imaging modalities are only moderately sensitive for detection of lymph node metastases in the pelvis, based on size criteria. However, a functional technique with a glucose analogue (18F-FDG) PET also performs poorly because of radioactive residues in the ureters and bladder.
Acknowledgments
We thank Daphne Glass for her invaluable help with the PET studies, Jayne
Morgan for help with CT, T. Krausz for his help with the histologic analysis,
and B. Puri for his statistical advice. We are grateful to Mary Crisp for her
secretarial assistance.
|
TopAbstractIntroductionMaterials and MethodsResultsDiscussionReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  P. Narayanan, T. Iyngkaran, S. A. Sohaib, R. H. Reznek, and A. G. Rockall Pearls and Pitfalls of MR Lymphography in Gynecologic Malignancy1 RadioGraphics, July 1, 2009; 29(4): 1057 - 1069. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 Y. Tsunoda, M. Ito, H. Fujii, H. Kuwano, and N. Saito Preoperative Diagnosis of Lymph Node Metastases of Colorectal Cancer by FDG-PET/CT Jpn. J. Clin. Oncol., May 1, 2008; 38(5): 347 - 353. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 E. Sala, S. Wakely, E. Senior, and D. Lomas MRI of Malignant Neoplasms of the Uterine Corpus and Cervix Am. J. Roentgenol., June 1, 2007; 188(6): 1577 - 1587. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H.-H. Chou, T.-C. Chang, T.-C. Yen, K.-K. Ng, S. Hsueh, S.-Y. Ma, C.-J. Chang, H.-J. Huang, A. Chao, T.-I Wu, et al. Low Value of [18F]-Fluoro-2-Deoxy-D-Glucose Positron Emission Tomography in Primary Staging of Early-Stage Cervical Cancer Before Radical Hysterectomy J. Clin. Oncol., January 1, 2006; 24(1): 123 - 128. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Sironi, A. Buda, M. Picchio, P. Perego, R. Moreni, A. Pellegrino, M. Colombo, C. Mangioni, C. Messa, and F. Fazio Lymph Node Metastasis in Patients with Clinical Early-Stage Cervical Cancer: Detection with Integrated FDG PET/CT Radiology, December 1, 2005; 238(1): 272 - 279. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
R. G. H. Beets-Tan and G. L. Beets Rectal Cancer: Review with Emphasis on MR Imaging Radiology, August 1, 2004; 232(2): 335 - 346. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
T.-C. Yen, K.-K. Ng, S.-Y. Ma, H.-H. Chou, C.-S. Tsai, S. Hsueh, T.-C. Chang, J.-H. Hong, L.-C. See, W.-J. Lin, et al. Value of Dual-Phase 2-Fluoro-2-Deoxy-D-Glucose Positron Emission Tomography in Cervical Cancer J. Clin. Oncol., October 1, 2003; 21(19): 3651 - 3658. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. Torizuka, T. Kanno, M. Futatsubashi, H. Okada, E. Yoshikawa, F. Nakamura, M. Takekuma, M. Maeda, and Y. Ouchi Imaging of Gynecologic Tumors: Comparison of 11C-Choline PET with 18F-FDG PET J. Nucl. Med., July 1, 2003; 44(7): 1051 - 1056. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. U. Herborn, T. C. Lauenstein, F. M. Vogt, R. B. Lauffer, J. F. Debatin, and S. G. Ruehm Interstitial MR Lymphography with MS-325: Characterization of Normal and Tumor-Invaded Lymph Nodes in a Rabbit Model Am. J. Roentgenol., December 1, 2002; 179(6): 1567 - 1572. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. A. Siegel, L. P. Adler, R. L. Wahl, A. Williams, N. DeSouza, and A. M. Peters FDG Positron Emission Tomography Detection of Pelvic Nodal Metastases in Gynecologic Cancer Am. J. Roentgenol., March 1, 2002; 178 (3): 762 - 764. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
