Patterns of Misinterpretation of Adnexal Masses on CT and MR in an Academic Radiology Department

Materials and methods 

The study population was identified by searching pathology reports from surgeries performed at Beth Israel Deaconess Medical Center, Boston from January 1, 2004, to May 9, 2005, containing the words “pelvic,” “pelvis,” “uterus,” “ovary/ies,” “tube(s),” “adnexa,” and/or “adnexal.” Patients in whom adnexal masses were reported by pathology were recorded. A total of 418 adnexal masses (from 325 women) with 650 sonographic studies, 145 CT studies, and 49 MR studies were reviewed in chronologic order by a gynecologic imager (DL, with 10 years of gynecologic imaging experience), with knowledge of histories but not of the written reports or histology. This patient population has been previously described with respect to sonographic and histologic findings 1, 2, 3. However, the CT and MR findings have not been previously described. Of these 325 women, 128 women had undergone pelvic CT studies (performed on four-slice, eight-slice, and 16-slice scanners) and MR studies (performed at 1.5 T) within 5 years of surgery, including 17 women undergoing both CT and MR studies. Up to three CT and three MR studies (the most recent studies with respect to surgery) within 5 years preceding surgery were included for each patient. This retrospective study was compliant with the Health Insurance Portability and Accountability Act and approved by our institutional review board, and informed consent was waived.

For each study, after the images were reviewed and a differential diagnosis was stated, the written report was then reviewed, and potential QA issues were noted. The imaging differential diagnosis was then correlated with the histologic diagnosis. If the reviewer felt that the histologic diagnosis did not correlate with the imaging appearance, the surgical report was reviewed. This was most important for assessment of endometriomas, peritoneal inclusion cysts, and hemorrhagic cysts that were drained at surgery and for diagnoses of adnexal torsion if necrosis was not evident at the time of histology. For cases with imaging findings that did not correlate with the surgical or pathologic findings, the images were reviewed by a pathologist with expertise in gynecologic imaging who reviewed images of the gross appearance (when available) and the histologic slides.

Each case was assigned a QA score of 0 (no QA issue), 1 (minor issue with minimal impact on clinical care), or 2 (major issue with potential impact on clinical care). These were subjective impressions based on information that a radiologist would have had available at the time of report interpretation. Lack of describing a lesion, lack of giving an appropriate diagnosis or differential diagnosis, and lack of appropriate imaging follow-up were all potential reasons for raising a QA issue.

After the identification of potential QA cases by the primary reviewer, a second review was performed by two abdominal imagers (IP, JK, with 6 and 10 years of abdominal imaging experience, respectively), who independently reviewed all the potential QA cases found by the primary reviewer, along with prior imaging, and assigned a score to each QA issue using the same scoring method.

The secondary reviewers gave each case a difficulty score for the complexity of making the diagnosis on the basis of the imaging available at the time of the imaging examination: 0 (very difficult diagnosis to make), 1 (difficult but possible to make the diagnosis, or 2 (diagnosis should be made).

Free-text comments were requested to describe reviewers' thoughts. If at least two of the three reviewers felt that a QA issue was present (a score of 1 or 2), and if both of the secondary reviewers gave difficulty scores of 1 or 2, the case was counted as a QA issue.

Indications and findings for the studies identified as having potential QA issues were tabulated. Reasons for the misinterpretation of each study were categorized by the subjective impression of the reviewers as to the likely etiology of the error.

To determine the population incidence of QA issues, the number of pelvic CT and MR examinations performed on women during the same time period was assessed.

Our institution is an academic medical center, with multiple subspecialists in abdominal imaging reading abdominal CT scans and MR specialists (not necessarily abdominal imagers) reading body MR cases. In addition, some pelvic CT and MR examinations performed for musculoskeletal indications are interpreted by musculoskeletal imagers. The abdominal imagers, MR specialists, and musculoskeletal imagers are not necessarily experts in women's imaging.

After data analysis was performed, all QA cases were entered into our online QA database. The results of QA cases were discussed in an abdominal imaging QA section meeting. In addition, the results of the study and images from select cases were discussed at grand rounds for both the obstetrics and gynecology and radiology departments.

Results 

The primary reviewer identified 30 potential QA issues in 28 women, 18 from CT studies and 12 from MR studies. QA issues were raised by the primary reviewer in 18 of 145 CT studies (12.4%) and 12 of 49 MR studies (24.5%). Two women were counted twice, one with issues raised on two CT studies and one with issues raised on CT and MR studies.

In 25 of 30 cases (83%), all three reviewers agreed that QA issues were present, and in 1 of 30 (3%) and 4 of 30 (13%) cases, one or the other secondary reviewer did not feel that a QA issue was present (Table 1). In 15 of 30 cases (50%), the two secondary reviewers each gave difficulty scores of 2, indicating that the diagnosis could have been made easily on the basis of the available images. One case, which received a difficulty score of 0 from both secondary reviewers, involved a malignancy that was reasonably interpreted as a physiologic cyst (Fig 1). This case was not counted in the overall statistics of QA issues.

Table 1. Computed Tomographic and MR QA Issues

	Patient	Modality	Indication	Age	Finding as Described in Written Report	QA Issue	QA Details	Median QA Score	Difficulty Scores	Final Diagnosis	Interval Between Imaging and Surgery∗
	1	CT	Von Hippel-Lindau syndrome, renal and pancreatic hemangioma	55	3.8-cm left adnexal cyst, “could be further assessed with pelvic US”	Inadequate follow-up recommendation	This was not a simple cyst on CT: US recommendation should be stronger in a postmenopausal woman with 3.8-cm complex cyst	1	2, 2	Dermoid	17 mo
	2	CT	Flank/back pain, microscopic hematuria	58	3.3-cm adnexal mass (left in findings, right in impression)	Inadequate description of location	Reversed right and left in impression	1	2, NA	Bilateral hydrosalpinx and paratubal cysts	1 mo
	3	CT	Left lower quadrant pain	49	12-cm simple-appearing adnexal cyst; recommended US or MR	Inadequate follow-up recommendation	MR should not be recommended when US is first-line study to assess internal characteristics of cyst	1	1, 2	Torsion	1 mo
	4	CT	Right lower quadrant pain	74	4.4-cm ovarian mass with mural calcification; “abnormal appearance of ovary”	No differential diagnosis given		1	2, 2	Simple cyst, calcified arterioles	3 wk
	5	CT	Right lower quadrant pain radiating to back	59	1.7-cm low-attenuation focus in ovary	Fat thought to be present on imaging, but no fat at pathology†		1	2, 2	Bilateral inclusion cysts and adhesions	1 mo
	6	CT	Left lower quadrant pain and fever	45	16-cm complex cystic structure, likely related to known left hydrosalpinx	Inadequate description of location	Only mentioned left side in bilateral process	1	2, 1	Bilateral tubo-ovarian abscess with intraovarian cysts and acute and chronic salpingitis	3 wk
	7a (first exam)	CT	Left flank pain, hydronephrosis	39	9-cm uterine fibroid, eccentric to right	Lack of correlation to prior studies	New mass since 5 years previously	2	2, 2	Non-Hodgkin lymphoma	5 mo
	7b (second exam)	CT	Left costovertebral angle pain	39	8.1-cm right pelvic mass, unchanged, likely fibroid	Lack of correlation to older studies	Compared to study 3 mo previously (7a) but not older studies	2	2, 1	Non-Hodgkin lymphoma	2 mo
	8	CT	Right lower quadrant pain, fever, abdominal distension	42	Tubular fluid–filled pelvic mass, likely hydrosalpinx, and pelvic free fluid	Lack of clinical correlation	No mention of infection in patient with fever, elevated white blood cell count	1	2, 2	Bilateral pyosalpinx, (right 10 cm, left 15 cm)	2 d
	9a (first exam)	CT	Gluteal pain	51	Gluteal abscess vs sarcoma with adjacent fat stranding	Missed lesion		2	2, 1	Dermoid (4 cm)	17 mo
	9b (second exam)	MR	Left gluteal pain and soft-tissue mass	51	Gluteal abscess	Missed lesion		2	2, 2	Dermoid (4 cm)	15 mo
	10	CT	Left lower quadrant pain	55	4.5-cm hemorrhagic cyst (right), 3.8-cm simple cyst (left); follow-up as clinically indicated	Inadequate follow-up recommendation	US indicated based on patient age and size of lesion	2	1, 2	Inclusion cyst (right), endometrioma (left)	17 mo
	11	CT	Lower abdominal pain	74	3.6-cm round dense mass in uterine fundus described as fibroid	Miscategorization		2	0, 1	Serous papillary carcinoma	6 mo
	12	CT	Abdominal pain, fever	48	8.7-cm and 4.7-cm bilateral degenerating exophytic fibroids	Miscategorization and lack of mention of nonvisualization of normal ovaries	Lack of visualization of normal ovaries not mentioned in report	2	2, 2	Bilateral serous papillary carcinoma	12 d
	13	CT	Known gastrointestinal stromal tumor, evaluate for chemotherapy response	77	None	Missed lesion		2	2, 2	Metastatic gastrointestinal stromal tumor (5 cm)	14 mo
	14	CT	Carcinoid tumor	35	“Prominent adnexa” unchanged since 2.5 mo previously	Inadequate follow-up recommendation	US was needed to characterize enlarged ovaries	2	2, 1	Metastatic carcinoid tumor (9 cm)	4 mo
	15	CT	Breast cancer and liver metastases	33	4.4-cm-thick walled cyst with asymmetric enhancement, suspicious for neoplasm	Inadequate differential diagnosis	Impression did not mention this cyst in patient with breast cancer	2	2, 2	Serous cystadenoma and paratubal cyst	2 mo
	16	CT	History of lymphoma	63	None	Missed lesion		2	2, 2	Serous cystadenoma (3.2 cm)	11 mo
	17	MR	Ovarian cyst seen on US at 20-wk gestational age	34	8.5-cm adnexal cyst, loss of signal on out-of-phase T1 and fat-saturated T1	Fat seen on MR but not at pathology†		1	2, 0	Epidermal cyst, no fat present	4 mo
	18	MR	Elevated liver function test results	51	None	Missed lesion, lack of comparison to prior studies	Known adnexal cyst not mentioned	1	2, 2	Serous cystadenoma (4.4 cm)	7 mo
	19	MR	Echogenic renal mass and ovarian cyst	57	1.5-cm right cyst, 2.0-cm left cyst, no nodularity or enhancement	Missed lesion	Solid Brenner tumor not mentioned	1	1, 1	Serous cystadenoma (right), Brenner tumor (left 2 cm)	3 mo
	20	MR	Back pain, known sacral/retroperitoneal mass	57	4.6-cm necrotic retroperitoneal mass, 1.3-cm endometrial lesion	Missed lesion	Necrotic ovarian mass with 7-mm solid nodule	1	2, 0	Metastatic endometrial cancer	2 mo
	21	MR	Adnexal cyst seen on US at 7-wk gestational age	34	7.1-cm heterogeneous adnexal mass with small area of fat signal intensity	Fat mentioned on MR but not seen at pathology†	Review of pathology showed a classic mucinous cystadenoma: the dumbbell-shaped lesion had fat outside the cystic components, misinterpreted as being within the cyst	2	2, 2	Mucinous cystadenoma	7 mo
	22	MR	Cervical cancer staging	49	8-cm adnexal cyst with smooth wall with mild enhancement and thin septation, corpus luteum, or cystadenoma	Inadequate differential diagnosis		1	2, 2	Hydrosalpinx, salpingitis	1 mo
	23	MR	Adnexal mass seen on US	64	3.1-cm heterogeneous mass, enhancement similar to myometrium, suggests fibroid	Miscategorization	No fibroid seen at surgery	2	2, 1	Paratubal cyst	12 d
	24	MR	Adnexal mass, fibroids	49	Enlarged cystic ovaries	Inadequate differential diagnosis	No mention of possible neoplasm	2	2, 2	Bilateral dermoids (right 7 cm, left 6 cm)	1 mo
	25	MR	Enlarged uterus on hysterosalpingogram	35	3.3-cm intramural/submucosal uterine lesion	Miscategorization	Adenomyosis called fibroid	1	1, 1	Adenomyosis	5 mo
	26	MR	Pelvic/rectal pain	59	Bilateral ovarian cysts, no pelvic abnormalities	Inadequate follow-up recommendation	Cysts in a postmenopausal woman should be further characterized by US; report gave no description of size of cysts	2	2, 2	Peritoneal inclusion cyst (right 5 cm, left 8 cm)	57 mo
	27	MR	Fibroids	49	3.9-cm right lesion with low-signal layering; normal left ovary	Inadequate description of location; inadequate comparison to prior studies	US had shown endometrioma on left	2	2, 1	Endometrioma (left 4 cm)	3 mo

CT, computed tomography; MR, magnetic resonance; NA, not applicable (the reviewer felt that the scoring system did not cover this issue); QA, quality assurance; US, ultrasound.

∗ Interval to surgery rounded to nearest month if >1 month. † Lack of fat confirmed by histopathologic review.

View Large Image Download to PowerPoint

Figure 1. A 45-year-old woman with colon cancer. Axial pelvic contrast-enhanced computed tomographic image, showing a simple-appearing right adnexal cyst with slightly thickened hyperattenuating wall (arrow). This was called a physiologic cyst, which was reasonable, because simple adnexal cysts are so common. However, the histologic diagnosis was metastatic colon cancer. This case was not ultimately categorized as a quality assurance case, because the primary reviewer flagged the case because of the missed diagnosis of malignancy, but all three reviewers felt that this could not be prospectively determined to be abnormal.

The categories of QA issues identified included observer error with missed lesions (n = 8, including three malignancies; Figure 2, Figure 3, Figure 4), adnexal lesions misidentified as leiomyomas (n = 6; Figure 5, Figure 6), fat described in the lesion but not seen histologically (n = 3; Fig 7), age or postmenopausal status of the patient not considered in the differential diagnosis or follow-up recommendations (n = 2; Fig 8), correlative ultrasound not recommended (n = 2), confusion of right and left sides (n = 2), and miscellaneous issues (n = 7). One case involved two QA issues, and two issues occurred twice for the same patients in two successive studies, 3 months apart. In one patient, an adnexal cyst was missed in an abdominal MR examination. Upon review of the MR images, the cyst was visible only on a scout view and at the edge of the field of view of a coronal T2-weighted single-shot fast spin-echo acquisition (Fig 2). In another patient, in whom a tailored MR examination for evaluation of the sacral plexus was performed, an adnexal mass was incompletely imaged only on the coronal T2-weighted fast spin-echo images and not detected on the initial interpretation of the study. In another patient, a right ovarian dermoid was missed on both CT and MR examinations tailored for the evaluation of hip pain (Fig 4).

View Large Image Download to PowerPoint

Figure 2. A 51-year-old woman with elevated liver function test results. (a) Coronal true fast imaging with steady-state precession (repetition time, 3.3 ms; echo time, 1.6 ms; matrix size, 256 × 256; slice thickness, 10 mm) magnetic resonance scout image and (b) coronal half-Fourier acquisition single-shot turbo spin-echo image (repetition time, 911 ms; echo time, 76 ms; matrix size, 256 × 205; slice thickness, 5 mm) show a 4.4-cm left adnexal cyst (arrow in a), which was not mentioned in the report. There is a single thin septation (arrowhead in b). The histologic diagnosis was serous cystadenoma.

View Large Image Download to PowerPoint

Figure 3. A 77-year-old woman with a gastrointestinal stromal tumor. Axial contrast-enhanced computed tomographic image shows a 5-cm right adnexal mass (arrow) with calcifications, which was not mentioned in the computed tomographic report. The histologic diagnosis was metastatic gastrointestinal stromal tumor.

View Large Image Download to PowerPoint

Figure 4. A 51-year-old woman with gluteal pain. Axial T1-weighted spin-echo (repetition time, 910 ms; echo time, 14 ms; matrix size, 512 × 512; slice thickness, 5 mm) magnetic resonance (a) and axial contrast-enhanced computed tomographic (b) images show a 4-cm fat-containing lesion (arrows), consistent with an ovarian dermoid, which was confirmed on histopathologic analysis. This lesion was not mentioned in either imaging report.

View Large Image Download to PowerPoint

Figure 5. A 48-year-old woman with abdominal pain and fever. Axial contrast-enhanced computed tomographic image shows a lobulated mass or two adjacent heterogeneous pelvic masses, one of which was called an exophytic fibroid and the other of which was thought to be the uterus, as well as a third right adnexal mass (arrows). The ovaries were not visualized. Histologic diagnosis was papillary serous carcinoma (metastatic).

View Large Image Download to PowerPoint

Figure 6. A 39-year-old woman with left flank pain and hydronephrosis. Axial computed tomographic images from two separate exams, 3 months apart, of an 8-cm, heterogeneous, eccentric pelvic mass (arrows), which was new since 5 years previously. On both exams, this was called a fibroid, but the histologic diagnosis was lymphoma.

View Large Image Download to PowerPoint

Figure 7. A 34-year-old woman at 20-week gestational age with an adnexal cyst seen on obstetric ultrasound. (a) Axial T2-weighted single-shot fast spin-echo (repetition time [TR], 727 ms; echo time [TE], 60 ms; matrix size, 256 × 160; slice thickness, 4 mm) magnetic resonance image shows an 8.5-cm right adnexal cyst with a small mural nodule (arrow) anteriorly. Axial gradient-echo (GRE) T1-weighted in-phase image (b) (TR, 205 ms; TE, 4.4 ms; matrix size, 256 × 160; slice thickness, 5 mm) shows high signal intensity in the nodule (arrow) compared to the low–signal intensity fluid in the cyst. Because of the loss of signal intensity within the nodule (arrow) on the T1-weighted opposed phase (TR, 205 ms; TE, 2.2 ms; matrix size, 256 × 160; slice thickness, 5 mm) (c) and fat-saturated GRE (TR, 180 ms; TE, 4.2 ms; matrix size, 256 × 160; slice thickness, 5 mm) (d) images, this cyst was called a dermoid. However, no fat was identified histologically, and the final diagnosis was epidermoid cyst.

View Large Image Download to PowerPoint

Figure 8. A 55-year-old woman with left lower quadrant pain. Axial contrast-enhanced computed tomographic image shows 4.5-cm and 3.8-cm adnexal masses (arrows). The report recommended further characterization by ultrasound “if clinically indicated,” but the ultrasound was not performed until 15 months later. In a postmenopausal patient, these findings are concerning, and an ultrasound is indicated on the basis of the imaging findings. No clinical correlation is needed. The histologic diagnosis was serous cysts.

During the time interval, women had a total of 11,194 pelvic CT scans and 603 pelvic MR scans performed at our institution, amounting to an overall incidence of QA issues of 0.15% (17 of 11,194) on CT imaging and 2.0% (12 of 603) on MR imaging.

Discussion 

There have been multiple studies assessing the accuracy of pelvic MR and CT studies in patients with known pelvic masses 4, 5, 6, 7. However, to our knowledge, there has not been a study assessing QA issues in patients who are not necessarily referred for adnexal masses.

In a 1997 review, Slanetz et al (8) found that 5% of women undergoing abdominopelvic or pelvic CT scans for indications other than ovarian carcinoma or known or suspected pelvic masses had incidental ovarian masses. This demonstrates that adnexal lesions are common.

Ultrasound is typically used to assess adnexal masses because it allows for the assessment of the internal characteristics of adnexal cysts. Although CT imaging can be used to assess for signs of malignancy and frequently can characterize fat and calcifications in dermoids (4), it has the limitation of poor soft-tissue discrimination (9). Therefore, when an indeterminate adnexal lesion is visualized on CT imaging, ultrasound is helpful to better characterize the lesion. In our study, not recommending further imaging evaluation with ultrasound was a frequent issue raised by the reviewers.

Our most common QA issue was that of missed adnexal lesions. Two of the missed lesions were dermoid cysts. Dermoids are a common ovarian neoplasm, typically being found in women of reproductive age (10). They are frequently found incidentally on physical examination or during surgery. Although the difficulty of detection of dermoids by sonography has been well documented 10, 11, to our knowledge the incidence of missed dermoids on CT and MR studies has not been described. It is reasonable that dermoids could be missed on cross-sectional imaging, because lesions containing fat can blend into the surrounding pelvic fat.

In addition to missed dermoids, we also had three missed malignancies: a Brenner tumor, a metastatic endometrial carcinoma, and a metastatic gastrointestinal stromal tumor. Two of these lesions were small, but the gastrointestinal stromal tumor was 5 cm. This emphasizes that the adnexal regions should be carefully assessed when interpreting cross-sectional pelvic imaging studies.

Another common reason for error was CT interpretation of a pelvic mass as a leiomyoma and thus failing to identify ovarian pathology. Leiomyomas typically enlarge the uterus, and thus on CT imaging, the uterus will have an abnormal contour or distortion of the endometrial cavity. An exophytic fibroid usually appears homogeneous, with similar attenuation to the uterus. However, leiomyomas may undergo cystic degeneration, appearing as a low attenuation or heterogenous attenuation uterine mass. Calcification is seen in 10% of leiomyomas on CT imaging (12). Although it is well recognized that a complex adnexal mass can be due to a leiomyoma mimicking adnexal malignancy 13, 14, the presence of a solid adnexal mass on CT imaging in a patient without documented leiomyomas on other imaging modalities (such as ultrasound or MR) should lead to further imaging to confirm the diagnosis. In our series of QA issues, three lesions (on four exams) that were called leiomyomas on CT studies were malignancies, one of which was lymphoma and two of which were papillary serous carcinomas.

A third common potential error was the identification of fat within lesions on CT and MR studies that did not contain fat histologically. Classic findings of dermoids on CT imaging include low attenuation of the cyst because of fat and calcifications due to tooth or bone formation. At times, a fluid-fluid or fat-fluid level may be present. On MR, classic signal characteristics of fat are the key to the diagnosis 15, 16, including high signal intensity on T1-weighted images, focal chemical shift artifacts 17, 18, 19, and loss of signal on fat-suppressed T1-weighted images. Fat within voxels can be seen by in-phase and out-of-phase imaging. Our series had three lesions for which fat was felt to be present on imaging that were not dermoids histologically. One of the reasons for this was felt to be a dumbbell-shaped mucinous cystadenoma with interdigitating fat in one case. A diagnosis of epidermoid cyst was rendered on histopathologic analysis in another case. The incidence of epidermoid cysts among surgically resected adnexal masses, thought to represent teratomas, may be as high as 17% (20). Although fat is not a characteristic feature of ovarian epidermoid cysts on pathology or cross-sectional imaging (21), cholesterol is frequently present in the cyst contents of these benign neoplasms (22) and may be responsible for the occasional presence of fat within these lesions on CT and MR imaging (23). It is also possible that small fat-containing foci within the lesions were not sampled by pathology.

Adnexal lesions in women aged >50 years constituted 15 of 30 of our QA cases (50%). Because menopausal status is frequently not given as part of the indication for a pelvic CT or MR study, the age of 50 (1 year younger than the average age of menopause in the United States, 51 years 24, 25) is a reasonable guideline to increase suspicion for a lesion when an adnexal cyst is seen. Although adnexal cysts are commonly found in postmenopausal women (26), on CT imaging, the internal characteristics cannot be completely described. Because cysts of any size can harbor malignancy, ultrasound is warranted in these patients to ensure that a benign-appearing cyst on a CT study is thin walled and anechoic on ultrasound (27). We use ultrasound for the initial characterization of adnexal masses that are incompletely characterized on CT imaging because it is low cost and readily available. MR is reserved for the characterization of masses that are incompletely described on ultrasound.

An important aspect of generating radiology reports is not just giving a description of the findings but also generating an impression and recommending appropriate follow-up. For example, in the breast, the Breast Imaging Reporting and Data System was developed in 1993 in response to variability in mammography reporting that resulted in ambiguity regarding the characteristics of breast masses, their estimated risk for malignancy, and the appropriate follow-up or intervention (28). Therefore we used the identification of lesions, giving appropriate differential diagnosis, and giving recommendations for appropriate follow-up when grading our QA issues.

As described previously, after the completion of our study, we entered each case into our online QA database for formal assessment. We presented our results at an abdominal imaging conference, as well as at two separate grand rounds, one for the obstetrics and gynecology department and one for the radiology department. This allowed for the dissemination of our findings to our practicing radiologists, trainees, and major referrers for gynecologic pathology. Future studies will need to be performed to assess how this intervention affects the interpretation of pelvic imaging studies in the future.

An important limitation of our study is that we used an enriched population, which included only patients who had undergone adnexal surgery. This allowed for a reference standard of histology and surgery but did not include all pelvic pathology imaged in the interval in individuals who did not have at least one ovary removed. Thus, this is an underestimation of QA issues, because lesions that did not have surgery were not found in our database search. This also allowed for a high index of suspicion for the primary reviewer to identify potential QA issues. However, use of the secondary reviewers allowed for agreement that QA issues were present in the majority of cases (29 of 30) identified by the primary reviewer. The identification of these errors and agreement among the three reviewers also highlights the value of peer review in potentially minimizing errors such as those identified in this study.

A second limitation is that the entire population was screened for QA issues by only the primary reviewer. This reviewer used a high index of suspicion for identifying adnexal lesions but could have missed some lesions that other reviewers would have identified.

A third limitation is the inclusion of adnexal lesions that were potentially missed because of incomplete or suboptimal evaluation on MR and CT examinations that used a protocol not tailored for evaluation of the pelvis. This approach, however, reflects clinical practice and illustrates the importance of careful review of all images in a given study, including the visualized portions of the pelvis in abdominal and pelvic MR and CT examinations, for potential incidental adnexal masses. The number of adnexal masses missed on MR examinations of other body parts that partially image the pelvis is unknown, and inclusion of these cases may have artificially increased the overall incidence of QA issues for pelvic MR examinations in our study. Similarly, the detection of adnexal masses may be more challenging in CT examinations depending on the imaging protocol (eg, no oral or intravenous contrast).

A fourth limitation is that our CT scanners were four-slice, eight-slice, and 16-slice scanners during the interval of imaging. Because of improvements in the image quality of CT and MR studies in recent years, including more recently operated patients might provide more up to date insight to current clinical practice.

In conclusion, errors on CT and MR studies regarding adnexal diagnosis in a highly enriched population of women undergoing adnexal surgery are common. In an academic radiology practice, pelvic CT and MR studies are often read by abdominal radiologists with different levels of expertise in gynecologic imaging. The most common clinically important issues were due to observer error, with missed lesions and lesions misidentified as leiomyomas. In addition, age and postmenopausal status should be considered for patients with adnexal masses. We recommend that presumed fibroids on CT imaging be confirmed with pelvic ultrasound and that care be taken to evaluate the pelvis with knowledge of the patient's menstrual history and age. These cases are presented as examples of QA issues in pelvic CT and MR imaging of adnexal masses to alert radiologists to common issues and decrease future errors.