Adnexal masses: development and preliminary validation of an MR imaging scoring system

To construct and undertake preliminary validation of a magnetic resonance (MR) imaging scoring system designed for use in pelvic MR imaging performed for characterization of adnexal masses that were indeterminate at ultrasonography (US).

The institutional ethics committee approved this retrospective study and granted a waiver of informed consent. The study population comprised 394 women who underwent MR imaging between January 1, 2008, and October 30, 2010, for characterization of 497 adnexal masses that were seen at US. Then, masses were chronologically divided into a training set (329 masses) and a validating set (168 masses). Two radiologists who were blinded to the clinical findings retrospectively evaluated MR imaging criteria for characterization of adnexal masses. In the training set, the positive likelihood ratio (PLR) of malignancy and κ values were calculated for each criterion. The reference standard was surgical pathologic findings or findings at imaging follow-up of at least 1 year. On the basis of the PLR and multivariate analysis, a five-category MR scoring system called the ADNEX MR SCORING system was created and was subsequently tested by six readers with the validating set.

There was almost perfect agreement (κ > 0.80) for each MR imaging feature except for grouped septa (κ = 0.558) and thickened regular septa (κ = 0.555). The classification was accurate in both the training set (area under the receiver operating characteristic [ROC] curve [AUC] = 0.981 for reader 1 and 0.961 for reader 2) and the validating set (AUC = 0.964 for reader 1 and 0.943 for reader 2). ROC curve analysis demonstrated that the optimal cutoff point was an ADNEX MR score of 3; an ADNEX MR score of 4 or higher was associated with malignancy with a sensitivity of 93.5% (58 of 62) and a specificity of 96.6% (258 of 267).

In this study, a reproducible and accurate MR imaging scoring system that has the potential to improve patient care was developed and tested. Multicenter prospective validation of the score is warranted.

http://radiology.rsna.org/lookup/suppl/doi:10.1148/radiol.13121161/-/DC1.