Development and Validation of an Interpretable Machine Learning Model for Predicting Early Pulmonary Metastasis Risk in Osteosarcoma

Authors:
Open Access
Publisher website
Google Scholar
Abstract
Introduction: Pulmonary metastasis in high‐grade conventional osteosarcoma remains a leading cause of treatment failure and mortality. Traditional monitoring methods are insufficient for early detection. This study aims to develop and validate an interpretable machine learning (ML) model using multi‐dimensional data from electronic medical records (EMR) to predict the early risk of pulmonary metastasis in high‐grade conventional osteosarcoma patients. Methods: This retrospective study included data from 522 high‐grade conventional osteosarcoma patients. After rigorous feature selection, 12 independent predictive factors were identified: prothrombin time (PT), international normalized ratio (INR), fibrinogen, globulin, prognostic nutritional index (PNI), eosinophil percentage, monocyte percentage, neutrophil percentage, monocyte count, maximum tumor diameter, gender, and amputation status. Eleven ML models were constructed and compared, with Shapley Additive Explanations (SHAP) analysis applied to enhance model interpretability and clinical transparency. Results: The gradient boosting model exhibited superior performance, achieving an area under the curve (AUC) of 0.891 in the training set and 0.742 in the independent test set. SHAP analysis revealed that tumor maximum diameter was the most influential predictor of pulmonary metastasis risk, while inflammation and coagulation‐related indicators also demonstrated significant contributions. Discussion: The gradient boosting model effectively predicts the early risk of pulmonary metastasis in high‐grade conventional osteosarcoma and provides interpretable risk factors. This model shows potential as a clinical decision support tool, facilitating personalized risk management and precision medicine, aiming to improve patient outcomes.
Funding Information
  • National Natural Science Foundation of China (82302690)

This publication has 42 references indexed in Scilit: