Predicting the Postoperative Recurrence Risk in Soft-Tissue Sarcomas of the Extremities and Trunk Using MRI-Based...

This study aims to develop a comprehensive nomogram for predicting the 3-year recurrence risk of patients with soft-tissue sarcoma (STS) undergoing surgical resection based on preoperative MRI images and clinical-radiological factors.

202 patients with STS of the extremities and trunk who had undergone surgical resection were included from two centers. We extracted tumor and peritumoral radiomics features from contrast-enhanced T1-weighted imaging (CE-T1WI) and fat-saturated T2-weighted imaging (FS-T2WI) sequences to construct corresponding models, and used pre-trained VGG11 and ResNet18 networks to build sequence-specific deep learning models. A clinical-radiological model was built using selected clinical-radiological features. Finally, deep learning, tumor and peritumoral radiomics, and clinical-radiological analysis results were integrated to construct a comprehensive nomogram for systematic evaluation and analysis from multiple perspectives.

Among all STS patients, the 3-year postoperative recurrence rate was 47.52% (96/202). The nomogram showed excellent predictive performance, with AUC values of 0.874(95% confidence interval [CI]: 0.761-0.987) and 0.822 (95% CI: 0.707-0.938) in internal and external validation sets, respectively; its concordance index for 3-year recurrence risk prediction was 0.746 and 0.690 in the two sets. Kaplan-Meier curves demonstrated significant prognostic differences in patient stratification across all cohorts (log-rank test, all p < 0.01).

The nomogram can predict the 3-year recurrence risk of patients, identify high-risk patients, and support personalized treatment.