Accuracy and Reproducibility of the O-RADS MRI Risk Stratification System Based on Enhanced Non-Dce MRI in the Assessment of Adnexal Masses

Liuhong Zhu
DOI: https://doi.org/10.1016/j.ejrad.2022.110670
IF: 4.531
2023-01-01
European Journal of Radiology
Abstract:Objective: Evaluation of the diagnostic performance and reproducibility of the Ovarian-Adnexal Reporting and Data System (O-RADS) Magnetic Resonance Imaging (MRI) risk stratification system based on enhanced non-dynamic contrast-enhanced (non-DCE) MRI in the diagnosis of adnexal masses. Methods: Patients who underwent conventional pelvic enhanced non-DCE MRI examination within one month prior to surgery formed the study population. Two experienced radiologists independently evaluated the images and assigned a score according to the O-RADS MRI risk stratification system. One of the radiologists reviewed the images and reassigned the scores after three months. Intra-and inter-observer agreement was evaluated with the k coefficient value. The adnexal masses that attained scores between 1 and 3 were considered benign, while those with scores of 4 or 5 were considered malignant. Analyses were conducted to determine the sensitivity, speci-ficity, positive and negative predictive values, and receiver operating characteristic (ROC) curve, which were then used for evaluating the diagnostic efficacy of the developed system based on enhanced non-DCE MRI scan. The reference standard was histology. Results: A total of 308 patients (mean age: 42.09 +/- 12.42 years, age range: 20-84 years) were enrolled in the study. Among the 362 adnexal masses from the included patients, there were 320 benign masses and 42 ma-lignant masses. In the case of three readers, there were no malignant tumors scored 1-2. The O-RADS MRI score >= 4 was associated with malignancy resulted in a good diagnostic efficacy with the areas under the curve (AUC) values of 0.918 (95 % CI, 0.864-0.972), 0.905 (95 % CI, 0.842-0.968), and 0.882 (95 % CI, 0.815-0.950), the sensitivity values of 90.5 % (95 % CI, 87.5-93.5 %), 85.7 % (95 % CI, 82.1-89.3 %), and 83.3 % (95 % CI, 79.5-87.2 %), and the specificity values of 93.1 % (95 % CI, 90.5-95.7 %), 95.3 % (95 % CI, 93.1-97.5 %), and 93.1 % (95 % CI, 90.5-95.7 %) obtained for the three readers, respectively. Excellent intra-observer agreement and inter-observer agreement were observed with the k values of 0.883 (95 % CI, 0.814-0.952) and 0.848 (95 % CI, 0.770-0.926), respectively.
What problem does this paper attempt to address?