Intratumor Heterogeneity Features Based on MRI Radiomics for Predicting Lung Metastasis Risk of Osteosarcoma

Motivation: Osteosarcoma is a heterogeneous tumor, and early prediction of lung metastasis is crucial for prognosis. We propose a predictive model based on intratumor heterogeneity (ITH). Goal(s): This study aims to develop an ITH index using MRI-derived radiomics features from osteosarcoma subregions and assess its predictive accuracy for lung metastasis within one year. Approach: A multicenter retrospective study of 320 osteosarcoma patients was conducted. MRI radiomics and ITH features were extracted, and multiple classifiers were tested. Results: The linear regression classifier showed the best performance, with AUC values of 0.875 and 0.843 in the internal and external test sets, respectively. Impact: This study presents an MRI-based ITH model, combined with clinical data, demonstrating significant potential for non-invasive lung metastasis risk assessment in osteosarcoma.