Detection

Endometrial cancer, the commonest gynecologic malignancy in the developed world, occurs predominantly after menopause with more than 90% of patients presenting over the age of 50 years. Transvaginal ultrasonography accurately evaluates the endometrial thickness and guides the requirement for endometrial biopsy, if the endometrial thickness is greater than 4 mm. Where available, the excellent soft tissue contrast of T2-weighted (T2-W) MR imaging provides the most accurate method of visualizing endometrial tumors. In some circumstances, the endometrium can be difficult to assess by T2-W MR imaging alone, particularly in cases where the endometrium is distorted by the presence of leiomyomas or adenomyosis. This visualization can be improved by the addition of DW-MR imaging (83% of tumors successfully identified compared with T2-W imaging alone, compared with 96% on fused DW-MR imaging and T2-W images).

Apparent diffusion coefficient (ADC) values derived from DW-MR imaging can be used to differentiate benign from malignant endometrial lesions. In 2 separate studies, the use of a cutoff ADC value proved useful in distinguishing benign from malignant lesions with an 85% to 87% sensitivity and 100% specificity. This is valuable in patients when preoperative endometrial sampling is not possible or when the endometrial pipelle does not adequately sample a tumor. Additionally, the detection of peritoneal dissemination improved with the combination of DW-MR imaging and conventional imaging (sensitivity, 84%; specificity, 91%), which enables preoperative treatment planning. MRS also has been explored in differentiating benign and malignant endometrial lesions largely through the assessment or quantification of the choline resonance. Lipid peaks may also be present in endometrial carcinomas, which are absent in benign endometrial lesions likely owing to cytosolic lipid droplets.

Staging

Federation Internationale Gynecologie Obstetrique (FIGO) staging remains surgically based after total abdominal hysterectomy, bilateral salpingooophorectomy, peritoneal washings, and retroperitoneal lymph node dissection. MR imaging is not currently included within the FIGO staging system. Both the National Cancer Institute in France and the European Society of Urogenital Radiology recommend preoperative MR imaging in the management of endometrial carcinoma, particularly those of high-risk histologic subtypes, because the information provided guides management and surgical planning.

Multiple studies have investigated the additional role of DCE-MR imaging for evaluating the depth of myometrial invasion in patients with endometrial carcinoma. Many found that the overall staging accuracy of T2-W imaging improved with the use of DCE-MR imaging, although this is improvement has not been seen in other series and is particularly important where fibroids and adenomyosis are present. A recent metaanalysis confirmed that DCE-MR imaging MR imaging is superior to T2-W imaging for both deep and superficial myometrial invasion, with a pooled specificity of DCE-MR imaging of 72% compared with 58% for T2-W imaging, but without a difference in sensitivity. Moreover, DCE-MR imaging improves confidence in staging accuracy in the presence of fibroids, adenomyosis, and in detecting cervical involvement thus distinguishing stage I and II disease. However, even with DCE-MR imaging, tumor extension into the cornua limits assessment of myometrial invasion and peritumoral inflammation may further overestimate it.

The accuracy of DW-MR imaging in assessing the depth of myometrial invasion ranges from 62% to 90% and so may be used where intravenous contrast medium is contraindicated ( Fig. 1 ). DW-MR imaging may also improve detection of drop metastases in the cervix or metastatic spread outside the uterus, such as in the adnexa or peritoneum ( Fig. 2 ). The conspicuity of lymph nodes improves with DW-MR imaging; however, the ability to differentiate benign and malignant lymph nodes remains controversial ( Fig. 3 ).

Local staging of endometrial cancer. Transverse T2-weighted image ( A ) shows a tumor within the endometrial cavity that has ill-defined margins with likely extension into the outer half of the myometrium ( arrow ). After intravenous gadolinium ( B ), there is marked peritumoral enhancement ( arrow ). The depth of myometrial invasion is clearly delineated on diffusion-weighted imaging (b = 1000 s/mm ² ; C, arrow ) and the corresponding ADC map ( D , arrow ) as less than 50% of the myometrial thickness, thus confirming a FIGO stage IA.

Pelvic spread of endometrial cancer. Transverse T2-weighted image ( A ) reveals a bulky tumor within the endometrial cavity ( arrow ) with corresponding restricted diffusion on diffusion-weighted imaging (DWI; b = 1000 s/mm ² ; B , arrow ). However, the left ovary also seems to be prominent on the T2-weighted image ( C , arrow ), and its marked restricted diffusion on DWI (b = 1000 s/mm ² ; D , arrow ) confirms the presence of an adnexal metastasis, upstaging the disease to FIGO stage IIIA.

Endometrial carcinosarcoma with metastatic spread. Transverse T2-weighted image ( A ) demonstrates an intermediate signal intensity tumor within the endometrial cavity and small bilateral obturator lymph nodes ( arrows ). The diffusion-weighted image ( B ; b = 1000 s/mm ² ) demonstrates restricted diffusion within the tumor and highlights a peritoneal metastasis abutting the sigmoid colon ( arrow ). Both lymph nodes seems to be restricted, although this does not definitively indicate metastatic involvement; the one on the left retains a fatty hilum.

An increase in the choline/water ratio on MRS with tumor stage has been reported with differentiation of FIGO stage I tumors from stage II and III tumors ( P = .029), but this is less accurate than MR imaging.

Characterization and Prognostication

The histologic grade and subtype of endometrial carcinoma on histology from biopsy is subject to sampling error. The ADC value has been interrogated to assess the aggressiveness of endometrial carcinoma. Rechichi and colleagues reported no correlation with tumor grade, depth of myometrial invasion, or the presence of lymph node involvement. The interquartile range of ADC indicative of tumor heterogeneity has also been shown to increase in deep myometrial invasion, lymphovascular space invasion, and lymph node metastases. ADC histogram analysis from the entire tumor volume indicated that the standard deviation, quartile, 75th, 90th, and 95th percentiles of ADC showed significant differences between all grades ( P ≤.03) and between high and low grades. The mean choline/water ratios do not differ between tumor grades.

Sarcomas do not seem to be different from carcinomas on conventional T2-W MR imaging. The vast majority of leiomyosarcomas arise de novo within the myometrium. Compared with their benign counterparts, leiomyosarcomas have an ill-defined and irregular margin, internal hemorrhage and necrosis, and demonstrate rapid growth. In addition, they often present with local and distant spread, including extrauterine tumor nodules, lymph node metastases, and liver, lung, and bone metastases. Both densely cellular leiomyomas and uterine sarcomas demonstrate restricted diffusion on DW-MR imaging, although the majority of benign leiomyomas do not demonstrate this marked restricted diffusion so that the mean ADC value of sarcomas is significantly lower ( Fig. 4 ). Endometrial stromal sarcomas are a rare entity and also show diffusion restriction, although also the evidence base is limited to case reports.

Leiomyosarcoma. Transverse T2-weighted image ( A ), diffusion-weighted imaging, ( B ) and apparent diffusion coefficient map ( C ) images through the mid pelvis showing a large, heterogeneous mass arising from the uterine body with lobulated margins ( arrow ). The irregular appearance and marked restricted diffusion in B and C with focal necrosis ( arrows ) is highly suggestive of leiomyosarcoma.

On MRS, uterine tumors characteristically show increased lipids ; lipid peaks were identified in 100% of uterine sarcomas, but not in the majority of benign leiomyomas. The choline peak is less reliable; it was demonstrated in all leiomyomas and 10 out of 12 sarcomas (the other 2 being markedly necrotic). The identification of lipids merits further investigation as a biomarker for differentiating benign leiomyomas from uterine sarcomas, particularly where management with hysterectomy versus uterine artery embolization or myomectomy is considered.

Monitoring Treatment Response

Because the majority of patients with endometrial carcinoma are treated and cured with surgery, there is a limited role for DCE- and DW-MR imaging in treatment response.

Follow-up and Detection of Recurrent Disease

Surgery is curative for the majority of patients with endometrial carcinoma, so it is important to identify patients at high risk of recurrence who merit imaging follow-up; these factors include advanced stage at diagnosis, high-grade disease, and lymphovascular space invasion. Of recurrences, 87% occur within 3 years of the initial diagnosis, most commonly with recurrent disease to the vaginal vault (42%) and lymph node metastases (46%). Computed tomography (CT) is commonly used to identify tumor recurrence; however, MR imaging is more sensitive in the identification of early vaginal vault recurrence. The use of DCE-MR imaging and DW-MR imaging aids recurrent tumor detection and allows differentiation from postradiotherapy changes. Nakamura and colleagues found that patients with a low minimum ADC of the primary tumor had a significantly lower disease-free survival than those with a high minimum ADC. Notably, the minimum ADC of the primary endometrial cancer was an independent predictive factor for disease recurrence in a multivariate analysis.

Detection

Cervical cancer occurs in a younger population of women (peak age incidence is 30 years). Cytologic screening has resulted in earlier diagnosis and surgery is the mainstay of management in these cases. Accurate depiction of the presence and extent of tumor is critical, particularly if fertility-sparing procedures are being considered. The performance of DW-MR imaging for tumor detection is equivalent to DCE-MR imaging ; the lack of requirement of an extrinsic contrast agent makes it preferable. The derived ADC value of malignant cervical tissues is significantly lower than that of normal tissue. This was shown in a metaanalysis of 13 cohort studies (645 tumor tissues and 504 normal tissues) and held true, regardless of data from 6 different scanner types within the analysis and of b-value combination. This is not surprising given the large differences in ADC between benign and malignant tissue, which is greater than the variability in the ADC calculation.

Where cervical cancer is detected incidentally on LLETZ/knife cone to treat dysplasia, imaging is required to identify residual disease or extension up the endocervical canal beyond the superior extent of the surgical specimen. In these small lesions, signal return may be improved 4 to 10 times by the use of endovaginal coils placed in close proximity to the cervix. In a study of 45 patients with small tumors (0.2 cm ³ ), one-third of whom had residual disease at fertility-sparing surgery, the sensitivity of detection using a 37-mm internal ring coil was 87% and specificity 80% and 90%, respectively, for 2 observers when DW-MR imaging was part of the evaluation ( Fig. 5 ). The use of DW sequences was particularly effective in improving sensitivity of tumor detection without loss of specificity after some form of local surgical excision (LLETZ or knife cone). A 10% to 30% increase in sensitivity of detected stage Ia and Ib1 tumors was achieved without a loss of specificity when an ADC threshold was applied and used in combination with T2-W images for disease detection. This technique, therefore, is helpful in selecting the appropriate operative procedure and in 1 study was shown to alter surgical management in 39% of cases. In small volume disease, DCE-MR imaging techniques have no place, because they are neither more sensitive nor more specific than T2-W imaging, nor is the information they provide adjunctive to the T2-W imaging data. MRS has insufficient spatial resolution in vivo, even with the increased signal available with an endovaginal technique. The most striking biochemical changes in malignant voxels are not only the increase in choline signals, but also the increased lipid resonances at 1.3 ppm (CH ₂ – resonances), 0.9 ppm (CH ₃ – resonances), and 2 ppm (NH-CH resonances). The technical issues with performing MRS in vivo, however, limit its translation into clinical practice.

Detection of cervical cancer using an endovaginal technique. T2-weighted coronal ( A ) and sagittal ( B ) images through the cervix showing a lobulated exophytic tumor on the left ( arrows ). The corresponding apparent diffusion coefficient (ADC) maps in coronal ( C ) and sagittal ( D ) planes show a restricted diffusion mass ( arrows ), but the heterogeneity of the ADC values within this adenocarcinoma are appreciated on these high spatial resolution images.

Staging

The use of DCE-MR imaging in distinguishing stage Ib1 from Ib2 tumors is equivalent to DW-MR imaging, and the need for an extrinsic contrast agent adds additional complexity and cost. A retrospective analysis of 152 cases for distinguishing stage 2b disease (parametrial invasion) using a whole-body technique showed that fused T2-W + DW-MR imaging had a higher specificity but not sensitivity than T2-W imaging for 2 readers: reader 1, 99% versus 88.7%; and reader 2, 96.5% versus 85.2%. Sensitivity for each reader remained between 67% and 75%, regardless of whether DW-MR imaging was added or not. The detection of parametrial extension may be boosted by the addition of tumor ADC data to the assessment of the parametrium itself ( Fig. 6 ) and in 117 patients treated surgically, tumor ADC and parametrial invasion on T2-W MR imaging were shown to be independent predictors of parametrial invasion on pathology. MRS has no role in this regard, because parametrial fat signals dominate and mask the recognition of smaller metabolites.

Staging a bulky cervical tumor. T2-weighted sagittal ( A ) and transverse ( B ) images through the cervix showing a homogenous, solid tumor anteriorly ( arrows ). The corresponding apparent diffusion coefficient (ADC) maps in sagittal ( C ) and transverse ( D ) planes show a restricted diffusion mass. The extension into the parametrium on the left is appreciated on the transverse ADC maps, but is not easily identifiable on the T2-weighted image ( arrows ).

Early data investigating ADC for detecting malignant nodes did not support its use, which is unsurprising because the parameter partly reflects the high cellular density present in lymphatic tissue. What is clear is that when nodes are enlarged the use of ADC is helpful; a significant difference in mean ADC has been shown in nodes greater than 5 mm (17 metastatic and 140 nonmetastatic). Therefore, when considered generally and without regard to size, ADC does not improve diagnostic accuracy; a prospective study of 68 patients showed a low sensitivity (25%–33%) on both a patient and regional node level for 2 observers; specificity was 83% to 97%. Of interest is the reproducibility of the detection of pelvic lymph nodes by DW-MR imaging, which has been shown to be equivalent to T2-W MR imaging, ranging between 42% and 65% for both methods. When compared with the current imaging gold standard, 18-fluorodeoxyglucose ( ¹⁸ FDG) PET-CT, ADC has been shown to be significantly different in PET-positive versus PET-negative nodes. However, the question remains as to whether the use of MR imaging is justified for delineating metastatic nodes if ¹⁸ FDG PET-CT can achieve a higher accuracy.

Characterization and Prognostication

Attempts have been made to derive imaging biomarkers that relate to histologic prognostic features. With DCE-MR imaging, the use of nonmodelled parameters such as the normalized relative signal increase and normalized area under the gadolinium enhancement curve are intrinsically more robust to obtain than modeled parameters and are potentially more useful. A retrospective study of 50 patients demonstrated the utility of a very simple parameter, the enhancing fraction (which represents the proportion of enhancing voxels within the tumor region) for indicating disease-free survival. Several pharmacokinetic parameters have also been associated positively both with locoregional control and progression-free survival; a prospectively collected cohort of 81 patients indicated that their prognostic impact was independent of tumor stage, volume, and lymph node status, was able to predict long-term locoregional control, and may be used as indicators of treatment failure.

More recently, the ADC has been shown to be an important prognostic indicator. Patients with lower mean or minimum ADC before hysterectomy had shorter disease-free survival than those with higher mean/minimum ADC values. This is supported by the relationship between ADC and histology: higher grade tumors have lower ADC values than their low-grade counterparts. Because the heterogeneity of tumors is increasingly recognized, histograms of voxel-wise analysis have proved more useful in distinguishing poorly differentiated from well differentiated tumors as well as adeno from squamous cancers. Intravoxel incoherent motion analyses in a small cohort of 16 patients distinguished well from poorly perfused tumors with good correlation between the intravoxel incoherent motion–derived perfusion parameter “f” (perfusion fraction) and K ^trans derived from DCE-MR imaging studies. More sophisticated analyses have attempted to combine ADC with total choline derived from MRS to classify tumors by type and grade, but the addition of the total choline measurement did not provide additional benefit.

Assessment of Treatment Response

Functional MR imaging is increasingly exploited during treatment to monitor response. Significant increases in DCE-MR imaging parameters (K ^trans and V _e ) have been shown within tumor, but not in normal muscle, 4 weeks into treatment or after 1 cycle of neoadjuvant chemotherapy that correlate with eventual tumor volume response rate. ADC values after 4 weeks of treatment also correlate with volume response and clinical response. Other studies have confirmed that the percentage ADC change after 1 month of chemoradiotherapy correlates positively with the percentage size reduction after 2 months of treatment. A larger study of 75 patients imaged after 2 and 4 weeks of commencement of chemoradiotherapy and at completion indicated a higher percentage increase in those confirmed as complete responders on tumor size criteria at 6 months than in those classified as partial responders. Similar early increases in ADC are evident in those treated with neoadjuvant chemotherapy. ADC increases correlate negatively with proliferating cell nuclear antigen and with cell density in those who respond indicating their relationship to cellular features of response that precede size reduction. Imaging very early (1 week) after initiation of chemoradiotherapy treatment with a combined DCE-MR imaging and DW-MR imaging approach in 16 patients showed that the changes in ADC correlated with size reduction on completion, although the changes in DCE did not, indicating a better sensitivity for DW-MR imaging as an early response biomarker and its future potential for directing therapeutic interventions at much earlier time points. At the other end of the spectrum, ADC measured at the end of treatment in a metaanalysis of 9 studies with 231 patients also showed that responders had higher ADC values and a greater change in ADC than nonresponders.

Follow-up and Detection of Disease Recurrence

The 4 factors most predictive of recurrence are size greater than 3 cm, adenocarcinoma, lymphovascular space invasion, and deep (outer third) stromal invasion. Between one-quarter and one-half of recurrences are asymptomatic. Because 64% to 87% of recurrences happen within 2 years and 89% to 98% within 5 years, there is a strong argument for 5-year surveillance. Morphologic pelvic MR imaging with T2-W sequences forms the mainstay in defining recurrent pelvic disease, whereas ¹⁸ FDG PET-CT remains the imaging modality of choice for estimating lymph nodes and distant disease, with a sensitivity and specificity of 96% and 95%, respectively. In patients with greater than 2 cm or node-positive disease treated surgically, a PET-CT at 3 months is recommended; in those treated with chemoradiotherapy, this should be delayed to 6 months to avoid false positives owing to postradiation effects on the images. MR imaging is reserved for symptomatic patients. However, in early cervical disease of less than 2 cm and node negative at the outset, follow-up imaging is warranted only if patients become symptomatic or if the physical examination is abnormal. T2-W MR imaging is the modality of choice for detecting locally recurrent disease in these cases.

The use of ultra-small iron oxide particles as a lymph node contrast agent enjoyed some early excitement and data from trials have affirmed its usefulness. Normal nodes take up the ultra-small iron oxide particles and become dark on T2* imaging, whereas metastatic involved nodes do not. However, these agents remain unlicensed for clinical use as their efficacy in phase III trials has not been proven. It is likely that their utility will be superseded with the development and availability of PET-MR.