Abstract
OBJECTIVES: To evaluate the effectiveness of habitat-based radiomics in differentiating early-stage serous borderline ovarian tumors (SBOTs) from serous malignant ovarian tumors (SMOTs), thereby enhancing diagnostic precision and treatment strategies. METHODS: We conducted a retrospective analysis of 210 patients with histopathologically confirmed SBOTs (n=95) and SMOTs (n=115) between December 2017 and February 2021. Multi-detector computed tomography images were obtained and analyzed using habitat-based radiomics, which segments tumors into distinct microenvironments based on Hounsfield Unit values. Clinical characteristics and imaging features were assessed, and predictive models were developed using logistic regression. Model performance was evaluated through receiver operating characteristic analysis, calibration curves, and decision curve analysis (DCA). RESULTS: The habitat-based models, the Habitat2 and the combined model, demonstrated high area under the curve values of 0.960 and 0.957 in the training set, with similar performance observed in the validation set. Solid components of tumors were identified as key differentiators, with only one radiomics feature from cystic regions retained in the final model. DCA indicated that habitat-based models provided significant clinical utility. CONCLUSIONS: Habitat-based radiomics model was developed and validated for accurately preoperative differentiation between SBOTs and SMOTs, emphasizing the importance of solid tumor regions for accurate diagnosis.