Predictive value of models based on MRI radiomics and clinical indicators for lymphovascular space invasion in endometrial cancer

BACKGROUND: Lymphovascular space invasion (LVSI), a prognostic indicator closely associated with tumour invasiveness, lymph node metastasis risk, and recurrence rate, is crucial in endometrial cancer (EC) staging; however, LVSI is currently diagnosed via postoperative pathology, highlighting the need for non-invasive diagnostic methods. This study aimed to investigate the predictive value of intratumoural and peritumoral magnetic resonance imaging (MRI) multiparametric radiomics combined with clinical indicators of LVSI in EC. METHODS: This retrospective analysis included 310 patients with EC who underwent preoperative MRI examinations at the Affiliated Hospital of Shandong Second Medical University (Centre A) and the First Clinical Medical College of Shandong Second Medical University (Centre B). The patients were divided into training (Centre A) and validation (Centre B) sets. Clinically independent risk factors and intratumoural and peritumoural radiomic characteristics were screened. Five models were constructed: clinical, peritumoural radiomics, intratumoural radiomics, combined intratumoural and peritumoural radiomics, and combined clinical, intratumoural, and peritumoural radiomics. A nomogram was constructed based on the optimal model. The diagnostic efficacy of the five models was evaluated using area under the curve. The accuracy of the model was evaluated using calibration curves, and the clinical value of the model was analysed using decision curve analysis. RESULTS: Logistic regression analysis identified CA125 and tumour length as independent risk factors for LVSI in EC. Among the five models, the combined clinical + intratumoural + peritumoural radiomics model performed slightly better than the other four models, with area under the curve values of 0.870 (95% CI: 0.821-0.919) for the training set and 0.818 (95% CI: 0.731-0.905) for the validation set. The calibration curve showed good consistency, and decision curve analysis suggested that the model had good clinical benefits. CONCLUSION: The combined clinical + intratumoural + peritumoural radiomics model based on clinical indicators and intratumoural and peritumoural multi-parametric MRI radiomics features demonstrated good diagnostic efficacy. This model provides a theoretical basis for preoperative evaluation of LVSI in EC.