Charis Bourgioti1, Konstantinos Chatoupis2, Evangelia Panourgias3, Chara Tzavara4, Kyrillos Sarris5, Alexandros Rodolakis6, Lia Angela Moulopoulos7. 1. Department of Radiology, Aretaieion Hospital, Medical School, University of Athens, 76 Vassilisis Sofias Ave., 11528, Athens, Greece. charisbourgioti@yahoo.com. 2. Department of Radiology, Aretaieion Hospital, Medical School, University of Athens, 76 Vassilisis Sofias Ave., 11528, Athens, Greece. kchatoupis@yahoo.gr. 3. Department of Radiology, Aretaieion Hospital, Medical School, University of Athens, 76 Vassilisis Sofias Ave., 11528, Athens, Greece. epanourgias@yahoo.com. 4. Department of Health, Epidemiology and Medical Statistics, Medical School, University of Athens, 25 Alexandroupoleos Str., 11527, Athens, Greece. htzavara@med.uoa.gr. 5. Radiation Oncology Department, Alexandra Hospital, 80 Vassilisis Sofias Ave, 11528, Athens, Greece. kyrilliss@yahoo.com. 6. Gynecologic Oncology Unit, 1st Department of Obstetrics and Gynecology, Alexandra Hospital, Medical School, University of Athens, 80 Vassilisis Sofias Ave., 11528, Athens, Greece. rodolaki@hol.gr. 7. Department of Radiology, Aretaieion Hospital, Medical School, University of Athens, 76 Vassilisis Sofias Ave., 11528, Athens, Greece. lia.angela1312@gmail.com.
Abstract
PURPOSE: To report discriminant MRI features between cervical and endometrial carcinomas and to design an MRI- scoring system, with the potential to predict the origin of uterine cancer (cervix or endometrium) in histologically indeterminate cases. MATERIALS AND METHODS: Dedicated pelvic MRIs of 77 patients with uterine tumors involving both cervix and corpus were retrospectively analyzed by two experts in female imaging. Seven MRI tumor characteristics were statistically tested for their discriminant ability for tumor origin compared to final histology: tumor location, perfusion pattern, rim enhancement, depth of myometrial invasion, cervical stromal integrity, intracavitary mass, and retained endometrial secretions. Kappa values were estimated to assess the levels of inter-rater reliability. On the basis of positive likelihood ratio values, an MRI-score was assigned. RESULTS: K value was excellent for most of the imaging criteria. Using ROC curve analysis, the estimated optimal cut-off for the MRI-scoring system was 4 with 96.6% sensitivity and 100% specificity. Using a ≥4 cut-off for cervical cancers and <4 for endometrial cancers, 97.4% of the patients were correctly classified. 2/58 patients with cervical cancer had MRI score <4 and none of the patients with endometrial cancer had MRI score >4. The area under curve of the MRI-scoring system was 0.99 (95% CI 0.98-1.00). When the MRI-score was applied to 20/77 patients with indeterminate initial biopsy and to 5/26 surgically treated patients with erroneous pre-op histology, all cases were correctly classified. CONCLUSION: The produced MRI-scoring system may be a reliable problem-solving tool for the differential diagnosis of cervical vs. endometrial cancer in cases of equivocal histology.
PURPOSE: To report discriminant MRI features between cervical and endometrial carcinomas and to design an MRI- scoring system, with the potential to predict the origin of uterine cancer (cervix or endometrium) in histologically indeterminate cases. MATERIALS AND METHODS: Dedicated pelvic MRIs of 77 patients with uterine tumors involving both cervix and corpus were retrospectively analyzed by two experts in female imaging. Seven MRI tumor characteristics were statistically tested for their discriminant ability for tumor origin compared to final histology: tumor location, perfusion pattern, rim enhancement, depth of myometrial invasion, cervical stromal integrity, intracavitary mass, and retained endometrial secretions. Kappa values were estimated to assess the levels of inter-rater reliability. On the basis of positive likelihood ratio values, an MRI-score was assigned. RESULTS: K value was excellent for most of the imaging criteria. Using ROC curve analysis, the estimated optimal cut-off for the MRI-scoring system was 4 with 96.6% sensitivity and 100% specificity. Using a ≥4 cut-off for cervical cancers and <4 for endometrial cancers, 97.4% of the patients were correctly classified. 2/58 patients with cervical cancer had MRI score <4 and none of the patients with endometrial cancer had MRI score >4. The area under curve of the MRI-scoring system was 0.99 (95% CI 0.98-1.00). When the MRI-score was applied to 20/77 patients with indeterminate initial biopsy and to 5/26 surgically treated patients with erroneous pre-op histology, all cases were correctly classified. CONCLUSION: The produced MRI-scoring system may be a reliable problem-solving tool for the differential diagnosis of cervical vs. endometrial cancer in cases of equivocal histology.
Entities:
Keywords:
Cancer; Cervix uteri; Endometrium; Immunohistochemistry; Magnetic resonance imaging