Imaging-Based Pre-Operative Differentiation of Ovarian Tumours—A Retrospective Cross-Sectional Study

<b>Objectives</b>: This study aimed to investigate the diagnostic performance of imaging-based biomarkers from computed tomography (CT) and magnetic resonance imaging (MRI) for prediction of malignant and borderline malignant ovarian tumours. <b>Methods</b>: 195 consecutive patients with suspected primary epithelial ovarian cancer were included from the retrospective "Prognostic and Diagnostic Added Value of Medical Imaging in Staging and Treatment Planning of Gynaecological Cancer" (PRODIGYN) study. The radiological stage, according to the International Federation of Gynaecology and Obstetrics system (rFIGO), magnetic resonance imaging (MRI)-based Ovarian-Adnexal Reporting and Data System (O-RADS-MRI) score, and the mean apparent diffusion coefficient (ADC_mean) were investigated for prediction of ovarian malignancy, with histopathology as reference. The same imaging biomarkers were applied to the borderline tumour cohort (n = 33) to predict malignant/adverse features, such as micro-invasion. <b>Results</b>: The rFIGO stage demonstrated high accuracy for ovarian malignancy, with an area under the curve (AUC) of 0.98 (95% confidence interval (CI) = 0.97-0.99). On lesion level, the sensitivity and specificity of the O-RADS-MRI score to predict ovarian malignancy, after adjusting for correlated data structure, was 1 (CI: 0.96-1) and 0.82 (CI: 0.70-0.90), respectively. The performance of ADC_mean to predict ovarian malignancy on lesion level was moderately high, with AUC = 0.78 (95% CI 0.68, 0.88). Discrimination of adverse features in borderline tumours was not improved. <b>Conclusions</b>: rFIGO and O-RADS-MRI showed excellent performance and outperformed ADC_mean as predictive tools for ovarian malignancy but could not predict adverse features in borderline tumours.