Preoperative contrast-enhanced CT-based deep learning radiomics model for distinguishing retroperitoneal lipomas and well‑differentiated liposarcomas

Abstract Objectives To assess the efficacy of a preoperative contrast-enhanced CT (CECT) –based deep learning radiomics nomogram (DLRN) for predicting murine double minute 2 (MDM2) gene amplification as a means of distinguishing between retroperitoneal well-differentiated liposarcomas (WDLPS) and lipomas. Methods This retrospective multi-centre study included 167 patients (training/external test cohort, 104/63) with MDM2-positive WDLPS or MDM2-negative lipomas. Clinical data and CECT features were independently measured and analysed by two radiologists. A clinical model, radiomics signature (RS), deep learning and radiomics signature (DLRS), and a DLRN incorporating radiomics and deep learning features were developed to differentiate between WDLPS and lipoma. Model utility was evaluated on the basis of the area under the receiver operating characteristic curve (AUC), accuracy, calibration curve, and decision curve analysis (DCA). Results In the external test cohort, the DLRN showed a higher AUC (0.861; 95% confidence interval, 0.737-0.985), accuracy (0.810), and net benefit than the RS, DLRS, and clinical model. The calibration curve analysis and DCA demonstrated that the nomogram exhibited good calibration and offered substantial clinical advantages. Conclusions The DLRN can accurately distinguish between WDLPS and retroperitoneal lipomas by predicting MDM2 amplification status. Clinical relevance statement The DLRN shows good diagnostic capability in predicting WDLPS and retroperitoneal lipomas preoperatively. It holds promise as an imaging biomarker for facilitating personalised management and precision medicine.