ABSTRACT
PURPOSE
To evaluate the accuracy of three-tesla (3T) magnetic resonance imaging (MRI) in determining deep myometrial invasion (DMI) in endometrial cancer (EC) cases and to reveal possible causes of discrepancy.
METHODS
Patients with EC who underwent preoperative pelvic MRI examination at a tertiary hospital were independently and blindly reviewed for DMI by two radiologists with differing levels of expertise. On MRI, the invasion of the endometrial mass into the myometrium was defined as superficial or deep (<50% or ≥50%) and was compared with the gold standard pathology reports, evaluated using the same criteria. Cases with discrepancies between MRI and pathology were re-evaluated, and the causes of the discrepancies were identified.
RESULTS
A total of 226 patients were included. The sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of MRI in detecting DMI were calculated as 77%, 93%, 84%, 88%, and 87%, respectively. In cases where radiology and pathology results were incompatible, expansile masses were found in 38%, myomas located near the lesion in 27%, masses located in the uterine horn in 21%, and in 14%, a specific cause explaining the error could not be identified.
CONCLUSION
MRI is the definitive diagnostic method for determining DMI. Awareness of these causes of discrepancy in MRI reporting can increase the value of the examination and contribute to patient management.
CLINICAL SIGNIFICANCE
This study highlights that 3T MRI has a high accuracy (87%) in detecting DMI in EC and demonstrates that analyzing misinterpretations can enhance reliability in patient management.
Main points
• Three-tesla magnetic resonance imaging (MRI) has a high accuracy (87%) in detecting deep myometrial invasion in endometrial cancer (EC).
• Identifying misinterpretations in imaging can improve diagnostic reliability and enhance patient management.
• MRI-based assessment of myometrial invasion can aid in preoperative planning, potentially reducing unnecessary extensive surgeries.
• Understanding the limitations and pitfalls of MRI interpretation is crucial for radiologists and clinicians in optimizing patient outcomes.
• A systematic approach to image analysis and error reduction can lead to better decision-making in EC management.
Endometrial cancer (EC) is the most common gynecological cancer in developed countries. The majority of cases occur in individuals aged between 60 and 70 years.1-3 EC is categorized into two histopathological subtypes: type 1 (80%–85%) and type 2 (10%–15%). Type 1 is estrogen-dependent and primarily affects younger women in the premenopausal or perimenopausal period. It consists of grade 1–2 endometrioid adenocarcinomas. In contrast, type 2 EC affects older women in the postmenopausal period. Histologically, it includes grade 3 endometrioid adenocarcinomas and other rare subtypes.4-7 Approximately 67% of patients present with early-stage disease, and at this stage, the 5-year survival rate is 81%.8 The prognosis depends on several factors, including histological type and grade, deep myometrial invasion (DMI), cervical invasion, and lymphovascular invasion.
The International Federation of Gynecology and Obstetrics (FIGO) classification system used for staging EC is a surgical staging procedure involving total hysterectomy with bilateral salpingo-oophorectomy, along with a thorough examination of the omentum, peritoneum, and intra-abdominal contents. Magnetic resonance imaging (MRI) is not included in the formal staging system for EC; however, it plays a critical role in treatment planning and determining the most appropriate initial therapy.3 According to the classification updated by FIGO in 2023, DMI–defined as infiltration of 50% or more of the myometrium–is one of the criteria for stage IB EC. Additionally, a new stage IC category was introduced for aggressive histological subtypes, provided the tumor is confined to a polyp or the endometrium. Accordingly, the prognostic significance of DMI, a key factor in EC treatment and outcomes, has increased.9-12 DMI is a predictor of extrauterine disease and lymph node metastasis. Lymph node metastasis occurs in 30% of DMI cases, compared with only 5% in cases of superficial myometrial invasion (SMI).13, 14
In this context, lymph node metastasis–the strongest predictor of recurrence–is associated with DMI. Lymphadenectomy may be considered for grade 3 and other high-risk EC cases. Additionally, in patients with grade 1–2 tumors, lymph node dissection may be considered if DMI is present.15 However, it has been shown that systematic lymphadenectomy does not improve overall survival in early-stage disease. Moreover, lymphadenectomy may lead to complications such as lymphedema and lymphocysts; therefore, it should be avoided in low-risk patients.16, 17 Identification of DMI through preoperative MRI assessment is thus clinically substantial for surgical planning.
In the literature, studies comparing MRI with intraoperative frozen section evaluation and postoperative pathological evaluation for determining DMI have reported high sensitivity, specificity, and accuracy for MRI, supporting its use as a superior diagnostic tool. MRI is the recommended imaging modality for preoperative radiological evaluation in EC.9, 16, 18-20 However, to our knowledge, no studies have specifically focused on the underlying causes of diagnostic errors in MRI.
One objective of this study is to evaluate the sensitivity, specificity, and accuracy of three-tesla (3T) MRI in detecting DMI by comparing its findings with postoperative pathological evaluations in cases of EC. Another goal is to identify the causes of discordant results by re-evaluating cases with incorrect radiological assessments through a secondary review.
Methods
Patient selection
This retrospective study was approved by the Ondokuz Mayıs University Clinical Research Ethics Committee (protocol number: 2022/450, date: 13.10.2022). The requirement for informed consent was waived due to the retrospective nature of the study. Patients with a pathological diagnosis of EC who underwent pelvic staging MRI between January 2016 and June 2022, and who subsequently underwent hysterectomy as part of surgery following the MRI, were eligible for inclusion (n = 240).
Cases with tumors extending beyond the uterus and serosa (n = 8) and those with artifacts hindering proper evaluation due to intestinal peristalsis or hip prosthesis (n = 5) were excluded. Additionally, cases in which the interval between MRI and surgery exceeded 1 month (n = 2) were also excluded. The remaining patients underwent surgery within 1 month after MRI. The study workflow is summarized in the flowchart (Figure 1).
Multiparametric magnetic resonance imaging protocol
To reduce artifacts caused by peristalsis, the patients were instructed to fast for 6 hours and to empty their bladders approximately half an hour before the examination. All examinations were performed using a 3T MRI scanner (Philips Ingenia, Netherlands) with a 16-channel torso coil. The patients were scanned in the supine position. High-resolution oblique axial and coronal images, sagittal T2-weighted images, axial T1-weighted images, axial diffusion-weighted images, and dynamic contrast-enhanced 3D T1-weighted images were obtained. For contrast-enhanced imaging, 0.1 mmol/kg of gadoteric acid was intravenously injected at a rate of 2 mL/s using a power injector, and images were acquired at 30, 60, 90, and 120 seconds. Imaging parameters are summarized in Table 1.
Image analysis
All images were independently analyzed by an abdominal radiologist with 20 years of experience and a general radiologist with 5 years of experience in abdominal/pelvic MRI interpretation. Both were blinded to the histopathological results. Preoperative MRI staging was performed according to the revised 2023 FIGO criteria. The degree of myometrial invasion was primarily assessed on high-resolution T2-weighted images obtained at a specific angle. Correlation was then made with non-angled diffusion-weighted imaging and contrast-enhanced series. The component of the endometrial lesion invading the myometrium was measured in millimeters and normalized to the myometrial thickness (Figure 2). It was documented whether myometrial invasion was <50% or ≥50%. During comparison of MRI results with histopathological findings, cases with discordant assessments were re-evaluated jointly by the two radiologists, and reasons for discrepancies were identified.
Histopathological examination
All pathological examinations were performed by an expert gynecological pathologist. Diagnoses were determined according to the 2014 World Health Organization classification of tumors of the uterine corpus. In the hysterectomy specimens, the depth of tumor invasion into the myometrium and the corresponding myometrial thickness were measured microscopically in a manner consistent with MRI assessment. DMI was considered present when the depth of invasion was ≥50%.
Statistical analysis
Statistical analyses were performed using the Statistical Package for the Social Sciences version 22.0 (IBM, New York, USA). Descriptive statistics were expressed as numbers and percentages for categorical variables and as mean, standard deviation, minimum, and maximum for continuous variables.
The chi-square test was used to assess differences between categorical variables. The sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy of MRI in detecting the depth of myometrial invasion were calculated. A P value of <0.05 was considered statistically significant.
Results
A total of 226 patients with EC were eligible for the study. The mean age of the patients was 60 ± 10 years (range: 36–94 years). The average size of mass lesions was 27 ± 14 mm (range: 10–92 mm). Following histopathological evaluation, 25% of patients had endometrial adenocarcinoma, endometrioid type, grade 1 (n = 58); 42% had grade 2 (n = 94); 11% had grade 3 (n = 24); and 22% had other less commonly observed histopathological diagnoses (n = 50).
According to EC types, the mean age of 152 patients (67%) with type 1 was 58 ± 9.5 years, and the mean age of 74 patients (33%) with type 2 was 64 ± 9.7 years.
In type 1 EC, 24.4% (n = 38) of patients had DMI and 75.6% (n = 114) had SMI. In type 2 EC, 56.6% (n = 43) had DMI and 43.4% (n = 33) had SMI. The difference between type 1 and type 2 with regard to DMI was statistically significant (P < 0.001). There was no statistically significant difference between type 1 and type 2 clinical classifications regarding discordant results between MRI and histopathology in detecting DMI (P = 0.307).
According to histopathological examination, DMI was detected in 77 patients (34%), whereas 149 patients (66%) had SMI. On MRI, 70 patients (31%) were found to have DMI, whereas 156 patients (69%) had SMI. In 11 patients with SMI on histopathological evaluation, MRI was interpreted as DMI. In 18 patients with DMI identified on histopathological evaluation, MRI was incorrectly interpreted as not having DMI, leading to understaging (Table 2). MRI had a sensitivity of 77%, specificity of 93%, PPV of 84%, NPV of 88%, and an accuracy of 87% in detecting DMI.
In 197 of 226 patients evaluated with MRI, the results were compatible with pathology in terms of DMI. In 29 patients with discrepant results, we investigated the causes of discrepancy through secondary retrospective evaluation (Table 3).
In 11 patients with incorrect staging on MRI, mass lesions were observed, causing thinning of the myometrium due to expansion. As a result, six of these patients were understaged and five were overstaged (Figure 3).
Eight patients had fibroids located in contact with or within 1 cm of the endometrial mass. Uterine fibroids disrupted the homogeneous structure of the myometrium, making it difficult to determine the extent of the lesion’s invasion. As a result, six patients were understaged and two were overstaged (Figure 4).
In 6 patients, the mass lesions were located in the uterine horn. The thinning of the myometrium at the level of the uterine horn made it difficult to determine the depth of myometrial invasion. As a result, three patients were understaged and three were overstaged (Figure 5).
In MRI evaluations of 4 patients where misclassification occurred, we could not identify any substantial factor that might have caused the discrepancy. We classified the cause in this group as “personal factors” (Figure 6).
We analyzed the frequency of previously identified factors leading to erroneous evaluations in the 197 cases where radiology and pathology results were consistent. Expansive masses were observed in 19 cases, masses near the uterus in 8 cases, and masses in the uterine horn in 5 cases. The frequencies of these findings were statistically significant in the group with erroneous evaluations (P < 0.001). The interobserver agreement in MRI evaluation was assessed using Cohen’s kappa coefficient, which was calculated as 0.831, indicating excellent agreement.
Discussion
In this study, which was conducted with EC cases, we compared MRI and histopathology results for determining DMI and successfully identified DMI with MRI in most patients, consistent with other studies in the literature.
In most cases where we misidentified myometrial invasion on MRI, there were mass lesions causing expansion. A mass lesion expanding within the uterine cavity results in thinning of the myometrium. In thin myometrium, it is expected that there will be a higher likelihood of measurement errors than in the normally thickened myometrium. In other studies, a frequently mentioned factor contributing to DMI evaluation errors is the presence of an expansile mass. Our review of the literature also identified postmenopausal atrophic myometrium as a potential error factor in some studies. In our cases, however, the error factors did not include isolated atrophic myometrium.
Lesions located in the uterine horn have also been identified as error factors in other studies. Myometrial thickness is low at the horn level, and its flexibility is limited. Therefore, even small masses can cause expansion and may hinder accurate evaluation. Subserosal or intramural myomas can disrupt the homogeneous structure and zonal anatomy of the myometrium, leading to possible mismeasurement of myometrial thickness and incorrect determination of DMI. Although some studies have cited the presence of adenomyosis as an error factor, it did not appear in the cases where we made incorrect staging.4, 6, 21-25
In a recent study, Liyanage et al.26 analyzed 235 women with EC and reported that MRI detected DMI with an overall accuracy of 86%, sensitivity of 72%, and specificity of 91%. They also showed that the presence of subserosal or intramural leiomyomas or adenomyosis reduced sensitivity while leaving specificity essentially unchanged.
In the literature, apart from the present study, there are numerous studies comparing preoperative MRI with postoperative histopathological assessment in cases of EC in terms of DMI. In these studies, the sensitivity of MRI in detecting DMI ranged from 50% to 91%, specificity from 72% to 94%, accuracy from 77% to 93%, PPV from 63% to 92%, and NPV from 81% to 92%.9, 16, 19, 27-31 In our study, although the sensitivity rate is lower than in some studies, we observed high specificity and accuracy rates.
In the 2023 revision of the FIGO staging system, stage 1 was further detailed, and additional subcategories were introduced. In this classification, both the presence of DMI and the aggressiveness of the histological subtype are taken into consideration.12 In our study, we found that aggressive histological subtypes did not negatively affect the accuracy of MRI assessments.
We demonstrated the high accuracy of MRI in determining myometrial invasion in EC cases, consistent with previous studies. We believe that considering the error factors identified in this study during MRI evaluation can enhance the reliability of the examination. Accordingly, recognizing MRI–pathology mismatches is essential: overstaging can expose patients to unnecessary lymphadenectomy, whereas understaging may delay adjuvant therapy, potentially affecting survival. Since DMI is an important marker for determining the risk of lymph node metastasis, it can assist in deciding whether lymphadenectomy should be performed prior to surgery. This approach may help prevent unnecessary lymphadenectomy and its associated complications.
In the literature, several studies have investigated the success of MRI in determining DMI, and as we mentioned in our article, various factors contributing to errors have been discussed. However, we did not find any other studies that specifically analyzed the causes of MRI errors and provided statistical data on these factors. Based on our literature review, our study includes one of the largest patient groups, which we consider a strength.
Our study has some limitations. First, it is a retrospective study. Second, in our MRI protocol, T2-weighted series were acquired in oblique axial and coronal views, but diffusion-weighted and dynamic contrast-enhanced series were not obtained in oblique planes. This may be considered a limitation, as it reduces the contribution of contrast-enhanced and diffusion-weighted sequences in determining DMI.
In conclusion, MRI has high accuracy in determining DMI in EC cases. However, erroneous interpretation of myometrial invasion may occur due to large expansile lesions, accompanying large myomas, and lesions located in the horn. We think that these factors should be mentioned in the radiology report when present, as they may act as misleading factors in evaluating DMI. This would enable more reliable treatment planning for the patient based on MRI findings.