T2-weighted MRI-based radiomics for preoperative prediction of homologous recombination deficiency status in high-grade serous ovarian carcinoma

Rationale and objectives: The aim of this study was to explore the feasibility of T2-weighted imaging (T2WI) - based radiomics in preoperative prediction of homologous recombination deficiency (HRD) status in patients with high-grade serous ovarian carcinoma (HGSOC), and to establish a non-invasive approach for guiding individualized treatment planning. Materials and methods A total of 133 patients with pathologically confirmed HGSOC who underwent MRI scan (T2WI) were retrospectively enrolled in this study. All patients were stratified into HRD-positive and HRD-negative group. These 133 patients (41 in HRD-negative group and 92 in HRD-positive group) were randomly allocated to a training set (n = 93) and a validation set (n = 40) at a 7:3 ratio. Three different models were constructed and evaluated: clinical characteristics only (clinical model); T2WI-radiomics features only (radiomics model); and a combination of them all (combined model). The predictive performance was evaluated using the receiver operating characteristic (ROC) analysis, and area under the curves (AUCs) were compared using the DeLong test. Calibration curves and decision curve analysis (DCA) were applied to validate the clinical-radiomic combined model. Results In the training set, the combined model achieved the best performance (AUC = 0.889; all P > 0.05). The radiomics model achieved predictive performance comparable to the clinical model (AUC of 0.762 vs. 0.790; P = 0.676). In the validation set, the combined model performed significantly better than the clinical or radiomics model (AUCs of 0.853 vs. 0.805 vs. 0.706; all P > 0.05). DCA demonstrated that the combined model achieved superior net benefit across a wide range of threshold probabilities compared with the clinical and radiomics models alone. Conclusions Our research demonstrated that the combined model achieved the optimal predictive performance. It is feasible to apply radiomics to predict HRD status of HGSOC patients, which can further assist clinicians in formulating individualized treatment strategies.