Deep Learning Radiomics Nomogram to Predict Lung Metastasis in Soft-Tissue Sarcoma: A Multi-Center Study

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence

Background:

• Soft tissue sarcoma (STS) is a rare malignancy.
• Predicting lung metastasis (LM) in STS is crucial for treatment planning.
• Current prediction methods have limitations.

Purpose of the Study:

• To develop and validate a deep learning radiomics nomogram (DLRN) for preoperative prediction of LM status in STS patients.
• To compare the DLRN's performance against clinical and radiomics-only models.

Main Methods:

• Retrospective analysis of 242 STS patients with magnetic resonance imaging (MRI).
• Feature selection using mRMR and LASSO algorithms.
• Machine learning classifiers (Logistic Regression, Decision Tree, Random Forest, SVM, AdaBoost) were trained and optimized with SMOTE for imbalanced data.
• A DLRN was constructed integrating clinical predictors and the best-performing radiomics signature (mRMR+LASSO+SVM+SMOTE).

Main Results:

• The DLRN achieved an Area Under the Curve (AUC) of 0.833 on external validation, outperforming the clinical model (AUC=0.664) and radiomics model (AUC=0.799).
• Calibration curves demonstrated good calibration efficiency.
• Decision curve analysis (DCA) indicated superior clinical applicability of the DLRN.

Conclusions:

• The DLRN is an accurate and efficient tool for predicting lung metastasis status in STS.
• This AI-driven approach enhances preoperative risk stratification for STS patients.
• The DLRN shows significant potential for improving clinical decision-making in STS management.