Pelvic lesions in patients with treated cervical carcinoma: efficacy of pharmacokinetic analysis of dynamic MR images in distinguishing recurrent tumors from benign conditions

Dynamic MR image series were analyzed with a pharmacokinetic two-compartment model. To preserve the spatial resolution of the dynamic MR images, the pharmacokinetic parameters were computed pixel by pixel, color coded, and superimposed on conventional MR images (pharmacokinetic mapping). The efficacies of pharmacokinetic mapping and conventional MR imaging in distinguishing between recurrent tumors and benign conditions in patients who have pelvic lesions after treatment of cervical carcinoma were compared. Twenty-one women with 24 suspected pelvic lesions and a history of treated cervical carcinoma (stages IB-IIIA) were included in this study. Patients had been treated before MR imaging with surgery or irradiation alone (eight patients) or a combination of the two (13 patients). Patients were referred because of our findings from CT examinations and/or clinical examinations. Of 24 suspected lesions, 17 were histologically verified as tumor recurrences and seven were classified as benign masses (histologically diagnosed as fibrosis and granulation tissue). T1- and T2-weighted spin-echo images were interpreted by three observers. During and after constant-rate infusion of gadopentetate dimeglumine, the kinetics of lesion response were determined with a strongly T1-weighted saturation recovery turbo-fast low-angle shot sequence. The signal-time curves for the suspected lesions were analyzed within the framework of a pharmacokinetic two-compartment model and displayed as color-coded images. The calculated pharmacokinetic parameters (amplitude [A] and tissue distribution time [t21]) were evaluated retrospectively to obtain optimal threshold values for differentiating malignant lesions from benign lesions. Analysis of the pharmacokinetic mapping data showed significantly shorter (p < .005) and stronger (p < .001) contrast medium enhancement of malignant lesions (t21, 24 sec; A, 1.5 arbitrary units) than of benign lesions (t21, 65 sec; A, 0.7), resulting in a sensitivity of 100%, a specificity of 88%, and an accuracy of 96%. Interpretation of the lesions on conventional T2-weighted MR images resulted in a sensitivity of 90%, a specificity of 38%, and an accuracy of 74%. Analysis of color-coded pharmacokinetic maps is more effective than conventional MR imaging in distinguishing between malignant and benign conditions in patients who have pelvic lesions after treatment of cervical carcinoma.