The accuracy of magnetic resonance (MR) imaging in staging invasive carcinoma of the cervix was determined retrospectively in 57 consecutive patients in whom the extent of disease was surgically confirmed. MR images were analyzed for (a) location and size of the primary tumor; (b) tumor extension to the uterine corpus, vagina, parametria, pelvic sidewall, bladder, or rectum; and (c) pelvic lymphadenopathy. The accuracy of MR imaging in determination of tumor location was 91% and for determination of tumor size within 0.5 cm, 70%. Its accuracy was 93% for vaginal extension and 88% for parametrial extension. Pelvic sidewall, bladder, and rectal involvement were accurately excluded in all patients, but the positive predictive values were 75%, 67%, and 100%, respectively. Overall, the accuracy of MR imaging in staging was 81%. MR imaging is valuable because it can accurately demonstrate tumor location, tumor size, degree of stromal penetration, and lower uterine segment involvement. It is also valuable for ruling out parametrial, pelvic sidewall, bladder, and rectal involvement.

Compared with CT, MRI offered significantly improved evaluation of tumor size, stromal invasion, and local and regional extent of disease in pre-treatment imaging for cervical cancer.

While the staging accuracy of both CT and MR imaging is only moderate, prediction of tumor resectability is excellent.

Magnetic resonance (MR) imaging for detection and characterization of ovarian masses was assessed in 33 patients with a total of 60 lesions. Lesions were characterized prospectively as benign or malignant by using T2-weighted MR images and unenhanced and gadolinium-enhanced T1-weighted MR images. MR imaging findings were compared with results of surgical laparotomy performed for staging of lesions. When malignancy was suspected, staging with MR imaging was performed. MR imaging demonstrated 57 of 60 (95%) surgically proved ovarian masses (34 of 36 were benign, 23 of 24 were malignant). Five significant primary criteria and four ancillary criteria for malignancy were established. For all MR pulse sequences combined, characterization of either type of lesion was correct in 84% of cases (48 of 57) when the five primary criteria were used and 95% (54 of 57) were correct when the four ancillary criteria were added. With gadolinium-enhanced images, correct characterization of malignant lesions increased from 56% to 78% with use of the five primary criteria and from 83% to 100% with use of both sets of criteria. Malignancies were correctly staged with MR imaging in 12 of 16 patients. Staging accuracy was 63% with unenhanced images and 75% with the addition of enhanced images.

The potential of magnetic resonance (MR) imaging in the detection of endometrial carcinoma and in the assessment of its extent was evaluated prospectively in 51 patients clinically suspected of having the disease. MR imaging findings were compared with the results of surgical-pathologic staging and lymph node sampling following hysterectomy. Histologic findings showed 45 patients to have endometrial carcinoma, three to have no residual tumor after dilatation and curettage, and three to have adenomatous hyperplasia of the endometrium. MR imaging demonstrated an endometrial abnormality in 43 of the 51 patients (84%). Endometrial carcinoma could not be differentiated from adenomatous hyperplasia or blood clots. Therefore, MR imaging was not specific for tumor detection, and histologic diagnosis remains essential. The overall accuracy of MR imaging in staging endometrial carcinoma was 92%; its overall accuracy in demonstrating the depth of myometrial invasion was 82%. Demonstration of lymphadenopathy and adnexal or peritoneal metastases by MR imaging was suboptimal.