| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Genitourinary Imaging |
1 From the Departments of Radiology (H.H., M.C., F.V.C., K.K., K.K.Y., G.S.), Epidemiology and Biostatistics (P.B.), and Gynecologic Oncology (C.B.P.), University of California-San Francisco, 505 Parnassus Ave, San Francisco, CA 94143-0628. Received January 12, 1999; revision requested March 5; revision received April 6; accepted May 4. Address reprint requests to H.H.
 |
Abstract |
|---|
 TOP Abstract IntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MATERIALS AND METHODS: Preoperative MR imaging of the pelvis was performed in 128 consecutive patients with clinically or ultrasonographically detected complex adnexal masses. Histopathologic examination demonstrated 187 masses, 96 of which were malignant. MR imaging studies were prospectively and independently reviewed by two radiologists, one of whom reevaluated the studies after a 6-month interval. The predictive value of MR imaging findings was determined with multivariate logistic regression analysis. The value of gadolinium enhancement was assessed by using receiver operating characteristic analysis. Inter- and intraobserver variabilities were assessed by using weighted 
statistics.
RESULTS: Gadolinium-enhanced MR imaging depicted 176 (94%) of 187 adnexal masses, with an overall accuracy for the diagnosis of malignancy of 93%. The MR imaging findings that were most predictive of malignancy were necrosis in a solid lesion (odds ratio, 107) and vegetations in a cystic lesion (odds ratio, 40). Use of gadolinium-based contrast material contributed significantly to lesion characterization. Interobserver (, 0.79–0.85) and intraobserver (, 0.84–0.86) agreement were excellent.
CONCLUSION: Gadolinium-enhanced MR imaging is highly accurate in the detection and characterization of complex adnexal masses, with excellent inter- and intraobserver agreement.
Index terms: Endometriosis, 852.3192 • Ovary, cysts, 852.311 • Ovary, neoplasms, 852.30, 852.323 • Pelvis, MR, 813.121411, 813.121415 • Receiver operating characteristic (ROC) curve • Teratoma, 852.313
|
Introduction |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
The main goal of imaging in the evaluation of an adnexal mass is the detection of malignancy (1). The standard of care for a suspected malignant adnexal mass is staging laparotomy with tumor debulking, which is performed preferably by an oncologic gynecologist (2). Ultrasonography (US) is the primary imaging modality for the assessment and characterization of adnexal masses, and the US features that indicate benignity are well established (3–12). However, the reported specificity of US for the diagnosis of benignity varies from 60% to 98%. In particular, as many as 20% of adnexal lesions in premenopausal women are classified as indeterminate by using US, even when they are interpreted in conjunction with clinical findings and CA-125 (ovarian cancer antigen) levels (13).
MR imaging has been shown to have potential in the characterization of adnexal masses: The results of two studies (14,15) demonstrated that MR imaging with gadolinium-based contrast material enhancement is superior to US. The tailored use of MR imaging in the evaluation of pelvic masses is also cost-effective (13,16). However, in these studies, a large number of patients were not examined and detailed multivariate analysis of MR imaging features was not performed. In addition, neither observer variability nor the role of intravenous gadolinium-based contrast material administration has been closely investigated. We undertook this study to further evaluate the accuracy of MR imaging in the detection and characterization of complex adnexal masses. In particular, we wished to determine which findings are most predictive of malignancy, assess the value of intravenous gadolinium-based contrast material administration, and assess inter- and intraobserver variabilities.
|
MATERIALS AND METHODS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MR imaging was considered to be clinically indicated for the evaluation of an adnexal lesion in all cases. None of the patients had contraindications to MR imaging. Of 174 patients, the 128 women who subsequently underwent surgery (laparoscopy or laparotomy) formed the study population. At surgery, 187 adnexal masses were found. The time interval between MR imaging and surgery was less than 1 month in 90 patients and between 1 and 2 months in 38 patients.
MR Imaging Technique
MR imaging was performed by using a 1.5-T MR imaging system (Signa; GE Medical Systems, Milwaukee, Wis) and either a body coil (n = 23) or pelvic phased-array coil (n = 105). Immediately before MR imaging, all patients were given 1 mg of intramuscular glucagon (Lilly, Indianapolis, Ind). None of the patients had contraindications to the administration of glucagon. The following images were obtained:
1. Sagittal T2-weighted fast spin-echo localizer images (repetition time msec/echo time msec, 4,000/102 [effective]).
2. Sagittal and transverse T2-weighted fast spin-echo images from the symphysis pubis to the aortic bifurcation (4,000–5,000/102 [effective]; echo-train length, eight; section thickness, 5 mm; intersection gap, 1 mm; field of view, 24 cm; four signals acquired and a 256 x 256 matrix used with the body coil and two signals acquired, a 512 x 256 matrix, and anterior and posterior spatial saturation bands used with the pelvic phased-array coil).
3. Transverse T1-weighted spin-echo images from the symphysis pubis to the aortic bifurcation (500–700/12; section thickness, 5–8 mm; intersection gap, 1–2 mm; field of view, 24 cm; one signal acquired; matrix, 256 x 192; use of respiratory compensation).
4. Optional transverse T1-weighted spin-echo images through any lesion with high T1 signal intensity by using the same parameters as those described above but with the addition of frequency-selective fat suppression. This sequence was performed when the supervising radiologist noted a lesion with high T1 signal intensity. Signal loss in the high-signal-intensity area of the lesion was considered to be diagnostic of fat in a cystic teratoma (17,18) (Fig 1).
|
|
6. Postcontrast transverse T1-weighted spin-echo images from the symphysis pubis to the aortic bifurcation after intravenous injection of 0.1 mmol/kg gadopentetate dimeglumine (Magnevist; Berlex Laboratories, Wayne, NJ) by using the same parameters as those used for precontrast T1-weighted imaging.
MR Image Analysis
The MR imaging findings were compared with the histopathologic findings, which were the standard of reference. Three separate MR image interpretations were recorded: (a) the formal report issued at the time of the MR imaging study by a senior radiologist (H.H.), (b) a separate independent retrospective reading by a junior radiologist (M.C.), and (c) a second reading by the senior radiologist at least 6 months after the first reading; this was done to evaluate intraobserver variability.
The senior radiologist had more than 10 years experience in pelvic MR imaging. The junior radiologist had 1 year of experience in pelvic MR imaging. The initial reading by the senior radiologist was performed with knowledge of all available clinical and US findings to optimize patient care. For the single reading of the junior radiologist and the second reading of the senior radiologist, the readers were aware that the patients were being evaluated for a possible adnexal lesion, but they were unaware of the other clinical or radiologic findings. Data were entered on a standardized form that was developed for the study.
We used the following previously established primary MR imaging findings to diagnose adnexal mass malignancy (19): mass size larger than 4 cm, presence of masses bilaterally, mass predominantly solid, necrosis in a solid lesion, and cystic lesion with wall or septal thickness greater than 3 mm or papillary projections. In addition, the following previously established (19) secondary MR imaging findings were used to diagnose adnexal mass malignancy: ascites, peritoneal metastasis, and adenopathy. The readers assessed each of these specific MR imaging findings.
A lesion was classified as malignant when at least two primary criteria were present or one primary and one ancillary finding were present. Each reader rated the conspicuity of each finding on nonenhanced and contrast material–enhanced images on a 1–5 scale, where 1 was extremely conspicuous and 5 was barely perceptible. For each MR examination, the readers estimated the likelihood that each of these findings was present by using a 0%–100% scale, where 0% was definitely absent and 100% was definitely present. Each reader also gave an overall impression of the likelihood of malignancy by using the same 0%–100% rating scale. Nonenhanced T1- and T2-weighted images and contrast-enhanced T1-weighted images were evaluated separately for lesion detection and characterization; the senior reader performed this detailed review at the 6-month delayed reading only.
Statistical Analyses
Multivariate logistic regression analysis was used to assess the predictive value of the presence of the five primary and three ancillary MR imaging findings used to assess for malignancy. The analysis was performed by both readers and by using nonenhanced and contrast-enhanced images. The outcome variable was the presence of malignancy at histopathologic examination. For logistic regression analysis, these results were dichotomized so that scores of greater than 60% were rated as "finding present," whereas scores of 60% or less were rated as "finding absent"; the choice of 60% as a threshold was based on the review of receiver operating characteristic curve analysis. Masses with a presumptive diagnosis of mature teratoma were excluded from MR image feature analysis.
The findings that were found to be statistically significant by using univariate analysis were entered into multivariate models to gauge their independent predictive value and determine which combination of findings would be most predictive of malignancy. The results were expressed as odds ratios of malignancy when a specific MR imaging finding was rated as definitely present compared with when it was rated as definitely absent. Descriptive statistical values, including accuracy, sensitivity, specificity, positive predictive value, and negative predictive value, were determined for each MR imaging finding. The pooled results of both readers were used to analyze the pre- and postexamination probabilities and determine the rates of detection with the nonenhanced and contrast-enhanced sequences. The statistical significance of differences in accuracies, sensitivities, and specificities was determined by using the McNemar test.
The areas under the receiver operating characteristic curves (Az) obtained by each reader at interpretation of both nonenhanced and contrast-enhanced images, as well as the asymptotic CIs, were calculated, and the statistical significance of the differences in these values between readers were determined by using permutation tests. The interobserver agreement on MR image interpretation between the senior radiologist and junior radiologist and the intraobserver agreement of the senior radiologist were quantified by using weighted statistics; a value of less than 0.40 was considered to represent poor agreement; 0.40–0.80, good agreement; and greater than 0.80, excellent agreement. The analyses were carried out by using Statistical Analysis System (SAS Institute, Cary, NC), Strata (Computing Resource Center, Santa Monica, Calif), and S-Plus (StatSci, Seattle, Wash) software packages.
|
RESULTS |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
|
Eleven lesions—six benign and five malignant—were not detected on either the nonenhanced or contrast-enhanced images. The benign lesions that were not detected were three peritubal cysts, one endometrioma, one ovarian hyperthecosis, and one hydrosalpinx. All of the undetected benign lesions were smaller than 2 cm in diameter. The malignant lesions that were not detected were two bilateral clear cell ovarian carcinomas that were mistaken for bowel loops, two bilateral adnexal malignancies that were misinterpreted as single large unilateral masses (Fig 2), and one papillary serous adenocarcinoma (2 cm in maximum dimension).
|
|
|
|
|
|
ranged from 0.79 with the nonenhanced images to 0.85 with all the MR imaging sequences combined. Similarly, there was excellent intraobserver agreement (, 0.84–0.86).
|
|
|
|
|
|
|
|
DISCUSSION |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
MR imaging is considered to be a problem-solving technique in the assessment of adnexal masses. Gadolinium-enhanced MR imaging appears to be more accurate than US in the assessment of adnexal masses (14,15). Nonenhanced MR imaging, contrast-enhanced computed tomography, and US all have similar accuracy (15,20–22). The results of our prospective study in a large patient population confirm previous reports suggesting that MR imaging is helpful in the evaluation of adnexal pathologic entities (14,15,19). Our findings demonstrate that gadolinium-enhanced MR imaging has a high rate of depiction of both benign (93%) and malignant (95%) lesions. Potential problems in lesion detection with MR imaging include small (<2-cm) lesion size and occasional difficulty in determining whether a large adnexal mass is unilateral or bilateral. The results of our study of lesion characterization are similar to previously published data (14,15,19). It is noteworthy that the results of our analysis of pre- and postexamination probabilities (Table 3) showed that MR imaging is most useful in confirming malignancy in complex adnexal masses. The postexamination probability of malignancy after a positive contrast-enhanced MR imaging examination was 94%, whereas the postexamination probability of malignancy after a negative contrast-enhanced MR imaging examination was 19%, which we believe is relatively high.
Our study results showed that contrast-enhanced MR imaging is significantly more accurate in lesion characterization than nonenhanced MR imaging. The improved characterization is due to greater conspicuity of the critical MR imaging findings, including the presence of solid components (or vegetations) in a cystic lesion and necrosis in a solid lesion. It should be noted that cystic teratomas (ie, dermoid cysts) are an exception to the conclusion that solid components in a cystic lesion imply malignancy. The presence of fat in a cystic adnexal lesion is diagnostic of a cystic teratoma, even if solid components (eg, Rokitansky nodules) also are present (17,18). Therefore, we excluded fat-containing lesions from our MR image feature analysis. The demonstration of fat requires both standard and fat-suppressed T1-weighted imaging, because the latter helps to differentiate fat from blood products as a cause of the high T1 signal intensity (23–25). Interestingly, the results of our characterization analysis showed that T2-weighted imaging does not add a significant diagnostic benefit to contrast-enhanced T1-weighted imaging. However, T2-weighted images may be helpful in the characterization of ovarian fibromas (26) (Fig 7), and we continue to obtain such images when evaluating adnexal masses.
|
|
In our study, although the combination of size larger than 4 cm and thickness of wall or septa greater than 3 mm also was significant in predicting malignancy, there was no improvement with the use of these criteria compared with the use of vegetation and necrosis as criteria. The results also indicated that lesion size, when combined with other imaging findings, does not further contribute to the prediction of malignancy (Fig 8). This appears to be discordant with the results of US-based studies, which have shown that lesion size larger than 4 cm substantially increases the likelihood of malignancy (28). The discrepancy is probably due to differences in patient selection. Most US studies include the evaluation of all adnexal masses and are often based on consecutive patients seen in primary care clinics. Such studies have a correspondingly high proportion of small benign lesions: Typically, 80% of the lesions are benign, and the average lesion is smaller than 4 cm. Our study included only patients referred from a gynecologic oncology clinic who had complex adnexal lesions, that is, adnexal masses that were not clearly benign after complete clinical and US assessment. It is likely that once small and clearly benign cystic lesions are excluded, size does not have a further role in lesion characterization. This finding suggests that lesion size can be used to aid in characterization at the initial US assessment but should not be used as a discriminator when an ultrasonographically suspicious or inconclusive lesion is referred for evaluation on the basis of MR imaging findings.
|
, 0.41) between two independent radiologists in the assessment of ovarian malignancy. In contrast, our study results demonstrated excellent interobserver agreement between two independent radiologists and excellent intraobserver agreement between the initial and 6-month delayed readings of the senior reader. The very high degree of intraobserver agreement is particularly noteworthy, because the second reading was performed without knowledge of the other clinical and radiologic findings, which were known at the initial reading. The lack of a difference between the two readings, even in the absence of clinical and US data at the second reading, indicates that the MR imaging findings used for characterization are objective and robust criteria. This further confirms that MR imaging is a reliable diagnostic technique for the characterization of adnexal masses. A potential criticism of our study is that the patient selection was biased, because all the patients had adnexal masses that were considered to be of sufficient concern at clinical or US evaluation to merit assessment at the gynecologic oncology clinic. However, despite this selection bias, nearly half (91 of 187) of the adnexal masses resected were benign; this indicated a wide spectrum of pathologic entities in our study group. In addition, this approach reflects the real-life use of MR imaging, which is reserved as a problem-solving modality. MR imaging is not used as the primary imaging tool in the evaluation of suspected adnexal masses.
Imaging algorithms are an essential part of clinical practice guidelines. On the basis of established practice and a review of the literature (3–12), we recommend that US remain the primary imaging modality for the evaluation of a clinically suspected adnexal mass. When the results of US evaluation are indeterminate, MR imaging is a cost-effective next step (13). Gadolinium-enhanced MR imaging is highly accurate in the detection and characterization of complex adnexal masses, with excellent inter- and intraobserver agreement. Gadolinium-based contrast material administration is recommended because it increases the conspicuity of findings that are predictive of malignancy. Our study results further validate the described primary and secondary MR imaging findings of adnexal malignancy (19) and show that MR imaging can be recommended as a reliable and reproducible modality in the assessment of complex adnexal masses.
|
Footnotes |
|---|
|
References |
|---|
|
TOPAbstractIntroductionMATERIALS AND METHODSRESULTSDISCUSSIONReferences |
|---|
This article has been cited by other articles:
|
|
 |
|
  I. Thomassin-Naggara, M. Bazot, E. Darai, P. Callard, J. Thomassin, and C. A. Cuenod Epithelial Ovarian Tumors: Value of Dynamic Contrast-enhanced MR Imaging and Correlation with Tumor Angiogenesis Radiology, July 1, 2008; 248(1): 148 - 159. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S Ghattamaneni, M J Weston, and J A Spencer Imaging in endometriosis Imaging, December 1, 2007; 19(4): 345 - 368. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. Adusumilli, H. K. Hussain, E. M. Caoili, W. J. Weadock, J. P. Murray, T. D. Johnson, Q. Chen, and B. Desjardins MRI of sonographically indeterminate adnexal masses. Am. J. Roentgenol., September 1, 2006; 187(3): 732 - 740. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 I. Imaoka, A. Wada, Y. Kaji, T. Hayashi, M. Hayashi, M. Matsuo, and K. Sugimura Developing an MR Imaging Strategy for Diagnosis of Ovarian Masses RadioGraphics, September 1, 2006; 26(5): 1431 - 1448. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 J A Spencer A multidisciplinary approach to ovarian cancer at diagnosis Br. J. Radiol., October 1, 2005; 78(Special_Issue_2): S94 - S102. [Full Text] [PDF]
|
|
|
|
|
|
K. Kinkel, Y. Lu, A. Mehdizade, M.-F. Pelte, and H. Hricak Indeterminate Ovarian Mass at US: Incremental Value of Second Imaging Test for Characterization--Meta-Analysis and Bayesian Analysis Radiology, July 1, 2005; 236(1): 85 - 94. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Saini, R. Dina, G. A. McIndoe, W. P. Soutter, P. Gishen, and N. M. deSouza Characterization of Adnexal Masses with MRI Am. J. Roentgenol., March 1, 2005; 184(3): 1004 - 1009. [Full Text] [PDF]
|
|
|
|
|
|
J A Spencer and M J Weston Imaging in endometriosis Imaging, June 1, 2003; 15(2): 63 - 71. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. A. A. Sohaib, A. Sahdev, P. V. Trappen, I. J. Jacobs, and R. H. Reznek Characterization of Adnexal Mass Lesions on MR Imaging Am. J. Roentgenol., May 1, 2003; 180(5): 1297 - 1304. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. Szklaruk, E. P. Tamm, H. Choi, and V. Varavithya MR Imaging of Common and Uncommon Large Pelvic Masses RadioGraphics, March 1, 2003; 23(2): 403 - 424. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. J. McClure, M. Atri, M. A. Haider, and J. Murphy Perineural Cysts Presenting as Complex Adnexal Cystic Masses on Transvaginal Sonography Am. J. Roentgenol., December 1, 2001; 177(6): 1313 - 1318. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
