July 2001, VOLUME 177

Recommend & Share

July 2001, Volume 177, Number 1

Genitourinary Imaging

Preoperative Diagnosis of Ovarian Tumors with MR Imaging
Comparison with Transvaginal Sonography, Positron Emission Tomography, and Histologic Findings

+ Affiliations:
1 Department of Diagnostic Radiology, University of Ulm, Robert-Koch-Str. 8, 89081 Ulm, Germany.

2 Department of Obstetrics and Gynecology, University of Ulm, Prittwitzstr. 43, 89081 Ulm, Germany.

3 Department of Nuclear Medicine, University of Ulm, 89081 Ulm, Germany.

Citation: American Journal of Roentgenology. 2001;177: 123-129. 10.2214/ajr.177.1.1770123

Next section

OBJECTIVE. Our study evaluated the diagnostic performance of MR imaging compared with that of transvaginal sonography and positron emission tomography (PET) in patients with clinically asymptomatic adnexal findings. An additional goal was to determine whether the combination of the three methods enhanced their diagnostic accuracy.

SUBJECTS AND METHODS. Included in the study were 103 women with suspicious adnexal findings on sonography. Patients underwent transvaginal sonography, MR imaging, and PET within 3 weeks of the initial sonography. For MR imaging, axial and sagittal T1-weighted gradient-echo sequences (unenhanced and enhanced) and T2-weighted turbo-spin-echo sequences were acquired. Transvaginal sonography was performed with a 7.5-MHz transducer head. For PET, a modern full-ring scanner was used. The results of diagnostic imaging techniques were first evaluated separately, and reviewers were blinded to the results of other methods. Finally, a second session resulted in a consensus diagnosis based on the findings of all three methods. Results of histology were considered the gold standard.

RESULTS. Histology revealed 12 malignant and 91 benign ovarian tumors. The following data were calculated for MR imaging, transvaginal sonography, PET, and consensus diagnosis: sensitivities, 83%, 92%, 58%, 92%; specificities, 84%, 59%, 78%, 84%; diagnostic accuracies, 83%, 63%, 76%, 85%, respectively. MR imaging, particularly with contrast-enhanced fat-saturated T1-weighted sequences, was found to correctly reveal dermoid and endometrial cysts. All three methods had false-negative findings with borderline tumors.

CONCLUSION. Transvaginal sonography is the diagnostic method of choice as a screening technique for ovarian processes. Suspicious findings on transvaginal sonography should be confirmed on MR imaging. If MR imaging confirms a dermoid or endometrial cyst, further diagnostic procedures may be unnecessary. In all other cases, a surgical evaluation must be considered.

Previous sectionNext section

Ovarian cancer has the highest mortality rate of all gynecologic malignant tumors [1, 2]. Nearly two thirds of all ovarian carcinomas have progressed to disease stage III or IV (Fédération Internationale de Gynécologie et d'Obstétrique) at the time of first diagnosis because they may remain clinically asymptomatic for extended periods [3, 4]. This fact underscores the importance of early detection of these tumors and of the correct determination of epithelial ovarian tumors of low malignant potential (so-called borderline tumors) [5, 6].

A number of cell surface antigens and serum proteins are produced by ovarian tumors and can be assayed with monoclonal antibodies. Some assays, such as CA 125 tumor-associated antigen, have been applied clinically as markers of disease status and may be helpful in the diagnosis of recurrent ovarian cancer [7]. In patients with early-stage disease, however, CA 125 is elevated in less than half of the patients [8]. As a screening tool, its sensitivity in the detection of early ovarian cancer is too low.

Sonography has been shown to be a sensitive, but relatively nonspecific method, leading to unnecessary surgical resection of many benign lesions [9]. The combination of sonomorphology and additional ovarian—blood flow measurements with Doppler sonography has been suggested to improve specificity [10]. However, numerous subsequent studies have failed to reproduce these results, whereas recently published data suggest that Doppler sonography does not increase specificity [11].

Besides clinical examination and transvaginal sonography, laparoscopy with histologic evaluation of biopsy material is the generally accepted gold standard for diagnosis of ovarian tumors [5, 12]. A disadvantage of laparoscopy is its invasiveness; patients with benign tumors are subjected to an unnecessary procedure. On the other hand, laparoscopic extirpation of malignant neoplasms is associated with the danger of rupture of the tumor's capsule with the potential for seeding of malignant cells in the peritoneum [13, 14].

Besides transvaginal sonography, which as a screening method can usually differentiate simple cysts from more complex lesions, CT is an option as an additional imaging technique. Because of its relatively poor soft-tissue contrast, the usefulness of CT in differentiating ovarian processes is limited. It is of particular use in cases of gynecologic malignancies to evaluate potential hematogenous, peritoneal, and lymphogenic spread [9].

Recently, MR imaging has been increasingly used because of its lack of radiation exposure and its good tissue contrast. On one hand, a prospective study has suggested that unenhanced MR imaging is inferior to transvaginal sonography in the diagnosis of adnexal lesions [15]. Enhanced MR imaging, on the other hand, has been shown to have higher diagnostic accuracy than transvaginal sonography [4, 16].

The role of positron emission tomography (PET) with fluorine-18 FDG for the diagnosis of ovarian tumors has remained controversial, with sensitivities between 83% and 86% and specificities between 54% and 86% [17,18,19,20].

Simultaneous evaluation of ovarian processes with all three diagnostic methods (transvaginal sonography, MR imaging, and FDG PET) has, to our knowledge, not yet been studied. The objective of our study was to evaluate transvaginal sonography, MR imaging, and FDG PET regarding their diagnostic accuracy in clinically asymptomatic adnexal lesions and to determine whether a combination of the three methods might be advantageous in the early detection of ovarian malignancies.

Subjects and Methods
Previous sectionNext section

In a 17-month period, 103 consecutive patients, 17-82 years old (average age, 46 years), underwent transvaginal sonography, MR imaging, and FDG PET. Each patient presented for a workup of a painless process of one ovary and had been referred to the department of gynecology as a result of suspicious adnexal findings on transvaginal diagnostic sonography. All pelvic masses were detected sonographically by routine gynecologic examinations of the referring physicians. Excluded from the study were those women in whom examinations unequivocally confirmed the malignant or benign nature of the process and those women in whom findings were bilateral. Also excluded were pregnant women. The study protocol was submitted to the institutional ethics commission and approved. Patients were informed regarding the character of the study, and all gave their written informed consent before examinations.

All patients underwent transvaginal sonography, MR imaging, and FDG PET in a 3-week period. The findings for each individual method were first evaluated by reviewers blinded to the results of the other methods. The evaluation of transvaginal sonographic findings was performed by two experienced gynecologists in consensus, whereas the evaluation of MR imaging findings was conducted by consensus of two radiologists experienced with the method. PET findings were evaluated by one nuclear medicine specialist with several years of experience with the method. In a second session, reviewers evaluated findings by reaching a consensus on the value of the three methods. Five imaging specialists attended. All results were compared with the findings of histology (or of cytology in one patient) and, in one patient, with results of sonographic follow-up (serous cysts).

Transvaginal and Doppler Sonography

Transvaginal sonography was performed by two experienced gynecologists using a 530 Combison unit with a 7.5-MHz transducer head (Kretztechnik, Zipf, Austria). Examination times averaged 30 min. The pelvis was examined in sagittal and coronal planes with the help of a transvaginal head. Patients with relatively large tumors also underwent transabdominal sonography with a 3.5-MHz transducer head. All normal and abnormal structures were documented and printed in black-and-white images. With findings showing abnormalities, examiners documented the lesion's exact size and location. Further differentiation was based on sonomorphology and was performed according to the morphology index of DePriest et al. [21] and the sonographic classification of ovarian tumors of Vera et al. [22] and Kawai et al. [23].

Thereafter, the Doppler gate was activated for color imaging of tumor vessels. Pulsed Doppler was used to interrogate each color signal detected, and the resistance index was calculated according to the following formula: resistance index = (peak systolic velocity—maximum end-diastolic velocity) / peak systolic velocity. If more than one tumor vessel was obtained, the lowest resistance index was used for further analysis. Criteria for suspicion of malignancy on transvaginal sonography were a pattern suggestive of malignancy (groups 9-12) in Vera et al. [22] and Kawai et al. [23], a DePriest score of 5 [21] or higher, or a resistance index equal to or less than 0.45 [24].

MR Imaging

All MR imaging examinations were performed with a 1.5-T Magnetom Vision unit (Siemens, Erlangen, Germany) with a body coil. For pelvic MR imaging, axial and sagittal T1-weighted gradient-echo sequences (TR/TE, 89.3/4.1) were acquired before and after application of contrast medium. Contrast-enhanced T1-weighted sequences with fat saturation were performed in 78 patients. IV gadolinium diethylenetriamine pentaacetic acid (Magnevist; Schering, Berlin, Germany) was administered at the usual dose of 0.1 mmol/kg of body weight. In addition, axial and sagittal T2-weighted turbo-spin-echo sequences (2800/138) were performed. Technical parameters were as follows: slice thickness, 6 mm; acquisitions, 2; field of view, 225 × 300 mm or 300 × 300 mm. The total examination period was approximately 45 min.

MR imaging findings were evaluated by a board-certified radiologist and a senior resident experienced in MR imaging. Divergent interpretations resulted in a second evaluation and a new diagnosis made with a consensus of both examiners. An evaluation sheet was used to document the exact localization and number of adnexal lesions. For each patient, the following characteristics were documented: wall thickness, the presence of septa, the contents of the lesion, the presence of cystic or solid components or of hemorrhage, and contrast medium—uptake behavior and signal pattern in the individual weighted sequences. Lesions were considered benign if one or more of the following criteria were met: exclusively cystic structures without any solid areas; diameter 4 cm or smaller; wall thickness less than 3 mm; presence of typical characteristics of a dermoid cyst or endometrioma [9, 25,26,27,28]. If one of these criteria was not fulfilled, the lesion was considered malignant. One exception was tumor size because larger benign ovarian processes are not uncommon, and the exclusive use of size as a criterion for malignancy has been shown to be relatively unspecific [9, 29]. Thus, lesions whose diameter exceeded 4 cm were also considered benign if in the opinion of the examiner, all other criteria for benign tumors were fulfilled. Additional criteria for malignancy included peritoneal, mesenteric, or omental disease manifestation and the presence of lymphomas or ascites [25]. In findings without evidence of malignancy, signal patterns in T1-weighted and T2-weighted sequences were used as the basis for diagnosis, such as in serous cysts, endometriomas, or teratomas. Criteria proposed in the literature were used as the basis for the diagnosis [9, 24,25,26,27].

Positron Emission Tomography (FDG PET)

PET examinations were performed with either the Ecat Exact HR+ or the Ecat 931/08/12 units (CTI-Siemens, Knoxville, TN). Both units acquire 63 or 15 transverse slices with a thickness of 2.46 or 6.75 mm, respectively, covering a field of view of 15.5 or 10.1 cm, respectively. All patients were fasting at the time of examination (12-hr fasting period with determination of fasting blood sugar level). About 50 min after IV injection of 222-555 MBq of 18F-marked FDG, static emission images were acquired from the liver to the inguinal region. Acquisition time was 12 min per bed position, with three bed positions usually required. To avoid artifacts and to reduce the concentration of radioactivity in the urinary tract, diuresis of patients was stimulated by means of IV injection of 20 mg of furosemide (Lasix; Hoechst, Frankfurt, Germany) and oral hydration. Images were reconstructed with Schmidlin's iterative reconstruction algorithm [30]. The resolution in each slice was 7 mm at the center of the field of view after iterative reconstruction (full width at half maximum). Evaluation of PET images was performed visually on transaxial, coronal, and sagittal slices; additional evaluation of three-dimensional rotational images (cine mode) was performed as needed. Lesions were considered suspicious for malignancy if their uptake of FDG equaled or exceeded that of the liver and the lesions were not localized in structures with physiologic FDG uptake, such as the ureter, urinary bladder, and bowel. FDG uptake patterns representing physiologic bowel activity were elongated curved sites of increased uptake projecting onto structures of the gastrointestinal tract.

After completion of the blinded review, data from all three diagnostic imaging methods were again evaluated simultaneously in a second session. Both the individual interpretations and the results of consensus evaluation were then compared, with the findings of histology as the gold standard.

Previous sectionNext section

Of the 103 patients, 12 had malignant and 91 had benign lesions. Most benign processes included serous cysts, endometrial cysts, and dermoids. Malignant lesions included five borderline tumors, three cystadenocarcinomas, three adenocarcinomas, and one instance of metastatic mammary carcinoma (Table 1). The frequency of malignancy was 11.7%.

TABLE 1 Summary of Histologic Findings of Adnexal Masses in 103 Women Studied

Transvaginal and Doppler Sonography

Correct diagnosis of malignant ovarian tumors was made in 11 of 12 patients on transvaginal sonography and Doppler velocimetry of tumor vessels. One borderline tumor escaped detection with these methods.

Benign lesions of the ovary were correctly identified on transvaginal sonography in 54 women on the basis of the previously mentioned classifications [21,22,23]. In 37 patients, benign ovarian processes were incorrectly diagnosed as malignant. Included in these false-positive lesions were cystadenomas, which were incorrectly considered malignant in 10 of 18 patients. Other false-positive findings on transvaginal sonography were five of 22 endometriomas, six of 18 follicular or corpus luteum cysts, four of eight cases of hydrosalpinx, four germ cell tumors, one adhesion, two cases of chronic salpingitis, one leiomyoma, one case of ovarian edema, one thecoma, and two nonovarian lesions (Table 1).

Individual evaluation of transvaginal sonographic and Doppler findings resulted in a sensitivity of 92% and a specificity of 59%. Positive predictive value was 23%, and the negative predictive value, 98%, with a diagnostic accuracy of 63% (Table 2).

TABLE 2 Statistic Analysis of All Three Methods and in Consensus

MR Imaging

Visual evaluation of MR imaging findings resulted in correct diagnosis of malignancy in 10 of 12 patients (Figs. 1A,1B,1C). The false-negative lesions included one borderline tumor and one cystadenocarcinoma (stage G3, pT3b, pN1). The borderline tumor was 4 cm in diameter and clearly circumscribed with tenuous contrast-enhanced septa. In addition, hemorrhagic areas in the tumor were visualized on T1-weighted sequences (Fig. 2A). This borderline tumor was considered benign at the second session, which assessed findings of all three methods, because both transvaginal sonography and FDG PET returned false-negative results (Figs. 2B and 2C). The cystadenocarcinoma, despite being larger than 4 cm in diameter, was also considered benign as a result of limited contrast medium uptake and relatively thin septa (<3 mm). The second consensus session, however, resulted in a correct diagnosis in this instance because both transvaginal sonography and FDG PET showed unequivocal signs of malignancy, with the result that the MR imaging findings were retrospectively reinterpreted to represent malignant disease (Table 1).

View larger version (157K)
Fig. 1A. 53-year-old woman with ovarian cancer. T1-weighted axial enhanced MR image shows cystic solid tumor of left ovary.

View larger version (107K)
Fig. 1B. 53-year-old woman with ovarian cancer. Corresponding transvaginal sonogram shows 30 × 40 mm solid portion against lateral wall of unilocular cyst.

View larger version (109K)
Fig. 1C. 53-year-old woman with ovarian cancer. Positron emission tomography scan shows increased fluorine-18 marked FDG uptake (arrow) in left adnexal region, which is typical for malignancy.

View larger version (178K)
Fig. 2A. 33-year-old woman with histologic confirmation of epithelial ovarian tumor of low malignant potential (so-called borderline tumor). T1-weighted axial enhanced MR image shows cystic tumor of right ovary without contrast enhancement of septa. MR diagnosis was endometrial cyst.

View larger version (121K)
Fig. 2B. 33-year-old woman with histologic confirmation of epithelial ovarian tumor of low malignant potential (so-called borderline tumor). Transvaginal sonogram shows multicystic lesion of right ovary. This lesion had no criteria typical for malignancy.

View larger version (110K)
Fig. 2C. 33-year-old woman with histologic confirmation of epithelial ovarian tumor of low malignant potential (so-called borderline tumor). Positron emission tomography scan did not show any fluorine-18 marked FDG uptake. Note physiologic FDG accumulation in urinary bladder. Thus, consensus diagnosis was endometrial cyst.

Correct diagnoses of benign ovarian processes were made in 76 of 91 patients on MR imaging. The 15 false-positive cases included five of 18 cystadenomas, four corpus luteum cysts, two cases of hydrosalpinx, one cholesterol granuloma, one case of chronic salpingitis, one case of edema of the ovary and, finally, one instance of normal adnexal findings in a patient with carcinoma of the rectum. This last patient, who had a solid tumor of the right side of the lesser pelvis, had a false-positive diagnosis in all three modalities, both individually and at the consensus evaluation session. Because no clear connection with intestinal structures had been visualized, the entity was incorrectly identified as an ovarian tumor. Even after the findings of histology were known, retrospective evaluation of MR images did not yield the definite diagnosis of a carcinoma of the rectum. Similarly, all three methods incorrectly diagnosed a luteal cyst. The morphology of the structure on both transvaginal sonography and MR imaging included solid portions and even displayed FDG uptake on PET, with the result that the consensus of the three methods suggested malignancy (Table 1).

Because of the presence of methemoglobin, endometrial cysts were characterized on MR imaging by homogeneous hyperintense signal patterns in T1-weighted sequences. In T2-weighted sequences, these cysts displayed hypointense signal patterns. Often a thick hypointense margin showed significant contrast medium uptake after IV administration of gadolinium dimeglumine [9]. As described by Ha et al. [27], the use of fat-saturated sequences was helpful in detecting smaller endometriomas.

Dermoid cysts were visualized on MR imaging as smoothly demarcated cystic lesions with solid portions. They showed a hyperintense signal pattern on T1-weighted sequences and a variable pattern on T2-weighted sequences. Fat-saturation sequences make it possible to differentiate a dermoid cyst from a hemorrhagic lesion, as other studies have reported [9, 26, 27] (Figs. 3A and 3B).

View larger version (186K)
Fig. 3A. 38-year-old woman with ovarian tumor on left side. T1-weighted unenhanced axial MR image shows well-circum-scribed inhomogeneous tumor (arrow) with partially bright signal.

View larger version (135K)
Fig. 3B. 38-year-old woman with ovarian tumor on left side. MR image after gadopentetate dimeglumine injection and fat suppression shows that tumor (arrow) has loss of signal intensity. Diagnosis was dermoid cyst, which was confirmed by histology.

Findings on MR imaging showed a sensitivity of 83%, a specificity of 84%, and a diagnostic accuracy of 83%. Positive predictive value was 40%, whereas the negative predictive value was 97% (Table 2).

Positron Emission Tomography (FDG PET)

Visual evaluation of FDG PET scans yielded the correct diagnosis of malignant ovarian tumors in seven of 12 patients. False-negative findings included three of five borderline tumors and two adenocarcinomas. In these cases, glucose metabolism in the tumor region was slightly increased, but not enough to permit definite assignment to either tumor or increased bowel activity (Table 1).

Findings in benign tumors were correctly identified as negative for malignancy in 71 patients. A characteristic appearance of benign cystic ovarian tumors was a photopenic defect without increased activity in either the margins or the interior of the cyst. False-positive findings were returned in 20 of 91 cases of benign processes: two cystadenomas, four endometriomas, five follicular or corpus luteum cysts, two cases of hydrosalpinx, three germ cell tumors, one cholesterol granuloma, one abscess, one the-coma, and one normal adnexal finding in a patient with carcinoma of the rectum (Table 1).

For individual evaluation, PET showed a sensitivity of 58%, a specificity of 78%, a positive predictive value of 26%, and a negative predictive value of 93%, with a diagnostic accuracy of 76% (Table 2).

Evaluation in Consensus

Consensus evaluation of all three methods resulted in correct diagnoses of malignancies in 11 of 12 patients. The exception was a borderline tumor that was interpreted as a false-negative finding in all three methods individually and in the consensus evaluation (Table 1 and Figs. 2A,2B,2C).

In 15 of 91 benign lesions, consensus evaluation revealed false-positive findings. Besides the case of rectal carcinoma with normal adnexal findings, findings of three cystadenomas were also interpreted as false-positive in both the individual and consensus evaluations. One endometrioma, three corpus luteum cysts, two cases of hydrosalpinx, one germ cell tumor, one cholesterol granuloma, one salpingitis, and one thecoma were considered suspicious for malignancy at the consensus evaluation after false-positive findings on MR imaging and transvaginal sonography or on MR imaging and PET, respectively. One of the endometrial cysts, despite un-remarkable transvaginal and MR imaging findings, displayed significant FDG uptake at PET scanning and was interpreted as potentially malignant. In one additional patient, the finding of an adnexal process described on histology as ovarian edema was incorrectly interpreted as positive for malignancy after false-positive transvaginal sonography and MR imaging findings, but negative PET findings (Table 1).

For the consensus evaluation, we calculated a sensitivity of 92%, a specificity of 84%, a positive predictive value of 44%, and a negative predictive value of 99%, with a diagnostic accuracy of 85% (Table 2).

Previous sectionNext section

Low survival rates among women with malignant processes of the ovary underscore the need for early detection and correct diagnosis of adnexal tumors [1]. The objective of our study was the separate evaluation of three modalities and the potential for improved diagnostic accuracy in combination. Our patients, with 12 malignant and 91 benign tumors with various histologic diagnoses, should be considered representative of the group of women with suspicious, though usually clinically unremarkable, adnexal findings.

Transabdominal sonography is a routinely performed modality when there is suspicion of abnormal adnexal findings. This method, however, may be limited in its efficacy because of overlying intestinal gas or interposed adipose tissue [31]. Transabdominal sonography may have an advantage over transvaginal sonography particularly in the case of large adnexal lesions [4]. The introduction of high-frequency transducer heads has made transvaginal sonography a reliable method, yielding high-resolution good-quality images. It is the first choice in the diagnostic step plan for workup of suspicious adnexal lesions [32].

Transvaginal sonography can usually differentiate between solid and cystic lesions. Its efficacy in revealing malignant versus benign ovarian processes, however, remains controversial, with sensitivities of 82-100% and specificities of 83-95% being reported in the literature [15, 16, 33]. These figures correspond to the findings in our study, which yielded a sensitivity of 92% but a specificity of only 59% for transvaginal sonography. Use of the scoring systems of Vera et al. [22], Kawai et al. [23], or DePriest et al. [21] did not significantly improve the accuracy of transvaginal sonographic findings in comparison with those of other studies. The use of these three different scoring systems may actually be one of the reasons for the somewhat lower specificity of transvaginal sonography in our study compared with other studies. In particular, the differentiation between malignancies, on the one hand, and dermoid cysts or endometrial cysts, on the other, proved problematic for transvaginal sonography. These findings contradict data published by Kombächer et al. [4], who reported a high detection rate on transvaginal sonography for dermoid cysts. This study, however, was conducted on a smaller patient group and without use of the previously mentioned scoring systems. In our study, the diagnosis of hydrosalpinx also proved problematic: here, transvaginal sonography allowed a correct diagnosis in only four of eight patients. This finding contradicts reports in the literature that describe superior results for transvaginal sonography in visualizing the convoluted sactosalpinx compared with results on MR imaging [4].

Adding color Doppler sonography to conventional analysis may improve sensitivity. However, because various thresholds for hemodynamic parameters have been proposed and a significant overlap between benign and malignant masses exists, the addition to conventional sonography of diagnostic information from Doppler sonography has been of limited value [11, 23, 24]. An explanation for the limited specificity and diagnostic accuracy of transvaginal and Doppler sonography in our study is the fact that those women in whom examination unequivocally confirmed the malignant or benign nature of the process and those women in whom findings were bilateral were excluded from the study.

Besides CT, the usefulness of which in the workup of adnexal lesions remains controversial, MR imaging is used increasingly as an additional cross-sectional imaging method [9, 34]. Advantages of the method include the high soft-tissue contrast, the different degrees of signal intensity, and absence of radiation exposure [9]. Unenhanced MR imaging is of only limited usefulness in the examination of adnexal lesions [15]. Reports in the literature differ with regard to the sensitivity and specificity of MR imaging in the differentiation of benign and malignant adnexal lesions, ranging from 85% to 95% and from 87% to 96% for sensitivity and specificity, respectively [16, 31, 34]. Our own calculated sensitivity (83%) and specificity (84%) are lower than figures published by Yamashita et al. [16], Thurner et al. [31], and the Radiological Diagnostic Oncology Group [35, 36]. An explanation for this fact may be our special study population. Those women in whom examination unequivocally confirmed the malignant or benign nature of the process and those in whom findings were bilateral were excluded from our study. Additionally, only 12 (11.7%) of 103 women had malignant tumors. In contrast, the Radiological Diagnostic Oncology Group reports the staging of advanced ovarian cancer [36]. In their earlier study, published in 1999, 118 (42.1%) of 280 patients had malignancies [35]. The French research group of Ghossain et al. [34] published figures for sensitivity and specificity of 85% and 87%, respectively. They, however, conducted their examinations with a 0.5-T MR imaging unit, so the results may not be directly comparable. Analogous to findings of other studies, MR imaging findings were particularly useful in revealing endometrial and dermoid cysts [9, 15, 25,26,27,28]. Stevens et al. [26] and Ha et al. [27] reported that both endometrial and dermoid cysts were best detected with a fat-saturation technique on T1-weighted sequences, as was done in most patients (n = 78) in our study, in which the number of false-negative or false-positive results was not influenced by the use of fat-saturated T1-weighted images. Nevertheless, for confidence in diagnosis, enhanced fat-saturated T1-weighted images were helpful.

For this reason, using MR imaging in the further workup of patients with suspicious adnexal findings on transvaginal sonography was proposed. In those patients in which MR imaging can reveal a lesion as either an endometrial cyst or a dermoid cyst, laparoscopy may be unnecessary. According to Yamashita et al. [16], higher diagnostic accuracy can be attained on MR imaging in mature cystic teratomas and endometriomas in comparison with that on transvaginal sonography. In other benign ovarian tumors, for example in cystadenomas, a specific diagnosis was possible in only a limited number of patients. This finding corresponds with reports in the literature that suggest that differentiation between cystadenomas and malignant tumors or cases of sactosalpingitis is often problematic [4]. Corresponding to our findings regarding a borderline tumor that was incorrectly diagnosed as a cystadenoma on both transvaginal sonography and MR imaging, an identical case was reported by Kombächer et al. [4]. This correspondence suggests that cystic tumors that are interpreted as benign on MR imaging may require further study to exclude the possibility of borderline tumors.

The sensitivity of FDG PET in the detection of ovarian carcinomas was only 58% in our study and is significantly less than figures reported in the literature, which have been in the range of 83-86% [17, 18, 20]. One reason may be the patient group recruited for our study, which consisted of selected asymptomatic women, only a small proportion of whom actually suffered from malignancies. The results of FDG PET showed that most malignant ovarian tumors possess an increased glucose metabolism and are visualized through an intense 18F-marked FDG uptake once lesions have reached a critical tumor cell mass. Early carcinomas and particularly borderline tumors presented problems because these tumors presumably lack the typical pattern of glucose uptake as a result of the small amount of transformed tissue. Our data yield a specificity of 78%, which corresponds with reports in the literature in the range of 54-86% [17, 18, 20]. However, published results differ with regard to the proportion of inflammatory changes and in the relatively small number of patients in the various study groups [17, 18, 20].

In three patients (one with abscess and two with sactosalpingitis), the well-known problem of the ability of FDG PET to reveal malignant versus inflammatory processes [37] led to false-positive findings. False-positive findings were also found with histologically distinct benign tumors, such as dermoid cysts and endometrial cysts. Increased FDG uptake in individual endometrial cysts has been reported in the literature [18, 20]. To our knowledge, nothing has been reported on the behavior of dermoid cysts of FDG PET scanning. Physiologic bowel activity may be associated with increased FDG uptake; a definite anatomic differentiation between bowel loops and adnexal structures may not always be possible. The usefulness of FDG PET in the preoperative determination of a lesion's malignancy is thus significantly limited.

The application of all three methods did not result in a statistically significant improvement in results. The sensitivity was not improved by combined application, in that a borderline tumor was considered a false-negative finding. Improved specificity resulted from a combination of transvaginal sonography and MR imaging (especially when fat-saturated sequences were used) with sonographically suspicious, but histologically confirmed, dermoids and endometrial cysts.

Our findings lead us to conclude that transvaginal sonography should remain the primary screening method for suspicious adnexal findings. Further workup of transvaginal sonographic findings suspicious for malignancy should begin with pelvic MR imaging with a fat-saturation technique. If MR imaging findings allow unequivocal diagnosis of a dermoid or endometrial cyst, laparoscopy may be safely deferred. All other adnexal lesions, despite absence of signs of malignancy on MR imaging, require laparoscopy. The presence of suspicious findings on MR imaging provides valuable additional information regarding the lesion's localization, extent, and relationship to adjacent anatomic structures, which is of importance in planning surgical treatment. FDG PET appears unsuitable as a routine diagnostic procedure and should be used only in selected patients in whom both transvaginal and MR imaging have failed to yield unequivocal findings.

Presented at the annual meeting of the American Roentgen Ray Society, Washington, DC, May 2000.

Address correspondence to A. Rieber.

Previous sectionNext section
1. Silverberg E, Lubera K. Cancer statistics. Cancer 1989; 39:3-20 [Google Scholar]
2. Barber HRK. Ovarian cancer. I. CA Cancer J Clin 1979; 29:341-351 [Google Scholar]
3. Nagell JR, Higgins RV, Donaldson E, et al. Transvaginal sonography as a screening method for ovarian cancer: a report of the first 1000 cases screened. Cancer 1990; 65:573-577 [Google Scholar]
4. Kombächer P, Hamm B, Becker R, Hese S, Weitzel H-K, Wolf K-J. Tumors of the adnexa: a comparison of magnetic resonance tomography, endosonography and the histological findings [in German]. Rofo Fortschr Geb Röntgenstr Neuen Bildgeb Verfahr 1992; 156:303-308 [Google Scholar]
5. Osmers R. Sonographic evaluation of ovarian mass and its therapeutical implications. Ultrasound Obstet Gynecol 1996; 175:428-434 [Google Scholar]
6. Sengupta PS, Shanks JH, Buckley CH, et al. Requirement for expert histopathological assessment of ovarian cancer and borderline tumors. Br J Cancer 2000; 82:760-762 [Google Scholar]
7. Teneriello MG, Park RC. Early detection of ovarian cancer. CA Cancer J Clin 1995; 45:71-87 [Google Scholar]
8. Jacobs I, Bast RC Jr. The CA 125 tumor-associated antigen: a review of the literature. Hum Reprod 1989; 4:1-12 [Google Scholar]
9. Outwater EK, Dunton C. Imaging of the ovary and adnexa: clinical issues and applications of MR imaging. Radiology 1995; 194:1-18 [Google Scholar]
10. Kurjak A, Schulman H, Sosic A, Zalud I, Hesham S. Transvaginal ultrasound, color flow and Doppler waveform of the postmenopausal adnexal mass. Obstet Gynecol 1992; 80:917-921 [Google Scholar]
11. Tekay A, Jouppila P. Controversies in assessment of ovarian tumors with transvaginal color Doppler ultrasound. Acta Obstet Gynecol Scand 1996; 75:316-329 [Google Scholar]
12. Cohen CJ, Jennings TS. Screening for ovarian cancer: the role of noninvasive imaging techniques. Am J Obstet Gynecol 1994; 170:1088-1094 [Google Scholar]
13. Seltzer V. Laparascopic surgery for ovarian lesions: potential pitfalls. Clin Obstet Ginecol 1993; 36:402-412 [Google Scholar]
14. Alvarez RD, Kilgore LC, Partridge EE, Austin JM, Shingleton HM. Staging ovarian cancer diagnosed during laparascopy: accuracy rather than immediacy. South Med J 1993; 86:1256-1258 [Google Scholar]
15. Jain KA, Freidman D, Pettinger TW, Allgappan R, Jeffrey RB, Sommer FG. Adnexal masses: comparison of specificity of endovaginal US and pelvic MR imaging. Radiology 1993; 186:697-704 [Google Scholar]
16. Yamashita Y, Torashima M, Hatanaka Y, Harada M, Higashida Y, Takahashi M. Adnexal masses: accuracy of characterization with transvaginal US and precontrast and postcontrast MR Imaging. Radiology 1995; 194:557-565 [Google Scholar]
17. Zimny M, Schröder W, Wolters S, Cremerius U, Rath W, Büll U. 18-fluordeoxyglucose PET in ovarian carcinoma: methodology and preliminary results [in German]. Nuklearmedizin 1997; 16:118-233 [Google Scholar]
18. Hübner KF, McDonald TW, Niethammer JG, Smith GT, Gould HR, Buonocore E. Assessment of primary and metastatic ovarian cancer by positron emission tomography (PET) using 2-18F-deoxyglucose (2-18F-FDG). Gynecol Oncol 1993; 51:197-204 [Google Scholar]
19. Karlan BY, Hawkins R, Hoh K, et al. Whole-body positron emission tomography with 2-F-18 Fluoro2-deoxy-D-glucose can detect recurrent ovarian carcinoma. Gynecol Oncol 1993; 51:175-181 [Google Scholar]
20. Römer W, Avril N, Dose J, et al. Metabolic characterization of ovarian tumors with positron-emission-tomography and F-18-fluordeoxyglucose [in German]. Rofo Fortschr Geb Röntgenstr Neuen Bildgeb Verfahr 1997; 166:62-68 [Google Scholar]
21. DePriest PD, Shenson D, Fried A, et al. A morphology index based on sonographic findings in ovarian cancer. Gynecol Oncol 1991; 51:7-11 [Google Scholar]
22. Vera MT, Okai T, Mukubo M, Kozuma S, Maeda T, Mizuno M. Ultrasonic evaluation and classification of ovarian tumors. Asia Oceania J Obstet Gynaecol 1986; 12:89-97 [Google Scholar]
23. Kawai M, Kano T, Kikkawa F, Maeda O, Oguchi H, Tomoda Y. Transvaginal Doppler ultrasound with color flow imaging in the diagnosis of ovarian cancer. Obstet Gynecol 1992; 79:163-167 [Google Scholar]
24. Alcazar JL, Ruiz-Perez ML, Errast T. Transvaginal color Doppler sonography in adnexal masses: Which parameter performs best? Ultrasound Obstet Gynecol 1996; 8:114-119 [Google Scholar]
25. Stevens SK, Hricak H, Stern I. Ovarian lesions: detection and characterization with gadolinium-enhanced MR-imaging at 1.5 T. Radiology 1991; 181:481-488 [Google Scholar]
26. Stevens SK, Hricak H, Campos Z. Teratoma versus cystic hemorrhagic adnexal lesions: differentiation with proton-selective fat-saturation MR imaging. Radiology 1993; 186:481-488 [Google Scholar]
27. Ha HK, Lim YT, Kim HS, Suh TS, Song HH, Kim SJ. Diagnosis of pelvic endometriosis: fat-suppressed T1-weighted vs conventional MR images. AJR 1994; 163:127-131 [Abstract] [Google Scholar]
28. Kier R, Smith RC, McCarthy SM. Value of lipid- and water-suppression MR images in distinguishing between blood and lipid with ovarian masses. AJR 1992; 158:321-325 [Abstract] [Google Scholar]
29. Granberg S. Relationship of macroscopic appearance to the histologic diagnosis of ovarian tumors. Clin Obstet Gynecol 1993; 36:363-374 [Google Scholar]
30. Schmidlin P, Doll J. Implementation of iterative reconstruction in positron emission tomography. Prog Clin Biol Res 1991; 363:59-67 [Google Scholar]
31. Thurner S, Hodler J, Baer S, Marincek B, Schulthess von GK. Gadolinium-DOTA enhanced MR Imaging of adnexal tumors. J Comput Assist Tomogr 1990; 14:939-949 [Google Scholar]
32. Bragg DG, Hricak H. Imaging in gynecologic malignancies. Cancer 1993; 71:1648-1651 [Google Scholar]
33. Sassone AM, Timor-Tritsch IE, Artner A. Transvaginal sonographic characterization of ovarian disease: evaluation of new scoring system to predict ovarian malignancy. Obstet Gynecol 1991; 78:70-76 [Google Scholar]
34. Ghossain MA, Buy J-N, Ligneres C, et al. Epithelial tumors of the ovary: comparison of MR and CT findings. Radiology 1991; 181:863-870 [Google Scholar]
35. Kurtz AB, Tsimikas JV, Tempany CM, et al. Diagnosis and staging of ovarian cancer: comparative values of Doppler and conventional US, CT and MR imaging correlated with surgery and histologic analysis—report of the radiology Diagnostic Oncology Group. Radiology 1999; 212:19-27 [Google Scholar]
36. Tempany CM, Zou KH, Silverman SG, et al. Staging of advanced ovarian cancer: comparison of imaging modalities—report from the Radiological Diagnostic Oncology Group. Radiology 2000; 215:761-767 [Google Scholar]
37. Kubota R, Yamada S, Kondo T, et al. Intratumoral distribution of fluorine-18-fluorodeoxyglucose in vivo: high accumulation in macrophages and granulation tissues studied by microautoradiography. J Nucl Med 1992; 33:1972-1980 [Google Scholar]

Recommended Articles

Preoperative Diagnosis of Ovarian Tumors with MR Imaging

Full Access,
American Journal of Roentgenology. 2010;194:311-321. 10.2214/AJR.09.3522
Abstract | Full Text | PDF (1159 KB) | PDF Plus (1185 KB) 
Full Access, , , ,
American Journal of Roentgenology. 2016;206:1351-1360. 10.2214/AJR.15.15199
Abstract | Full Text | PDF (1078 KB) | PDF Plus (1140 KB) 
Full Access, , , , ,
American Journal of Roentgenology. 2001;177:237-243. 10.2214/ajr.177.1.1770237
Abstract | Full Text | PDF (5426 KB) | PDF Plus (5517 KB) 
Full Access, , , ,
American Journal of Roentgenology. 2014;202:1366-1375. 10.2214/AJR.13.12023
Abstract | Full Text | PDF (940 KB) | PDF Plus (991 KB) 
Full Access, , , ,
American Journal of Roentgenology. 2003;180:1297-1304. 10.2214/ajr.180.5.1801297
Abstract | Full Text | PDF (6840 KB) | PDF Plus (811 KB) 
Full Access, , , , ,
American Journal of Roentgenology. 2010;194:330-336. 10.2214/AJR.09.3936
Abstract | Full Text | PDF (950 KB) | PDF Plus (924 KB)