Survival Prediction in Brain Metastasis Patients Treated with Stereotactic Radiosurgery: A Hybrid Machine Learning...

Area of Science:

• Oncology
• Radiotherapy
• Machine Learning
• Biostatistics

Background:

• Accurate survival prediction is critical for personalized treatment planning in brain metastasis patients undergoing stereotactic radiotherapy (SRT).
• Current prediction methods may lack the precision needed for optimal clinical decision-making.
• Developing advanced computational tools can significantly improve patient outcome management.

Purpose of the Study:

• To develop and evaluate a machine learning model for precise survival time estimation in brain metastasis patients treated with SRT.
• To compare the performance of multiple machine learning algorithms, including XGBoost, CatBoost, Random Forest, and Gradient Boosting.
• To identify the most effective algorithm for clinical application in predicting patient survival.

Main Methods:

• A hybrid machine learning approach was employed using real-world data from patients with brain metastasis treated with SRT.
• Four algorithms (XGBoost, CatBoost, Random Forest, Gradient Boosting) were compared within a meta-model framework.
• Model performance was rigorously assessed using Mean Squared Error (MSE), Mean Absolute Error (MAE), Mean Absolute Percentage Error (MAPE), and C-index.

Main Results:

• The XGBoost model demonstrated superior performance, achieving an MSE of 0.14%, MAE of 0.10%, and MAPE of 0.093%, with a 100% C-index.
• CatBoost showed moderate performance, while Gradient Boosting exhibited higher error rates (MSE: 6.99%, MAE: 21.04%, MAPE: 19.29%).
• Random Forest yielded the weakest results, with the highest error metrics (MSE: 14.39%, MAE: 30.23%, MAPE: 33.58%).

Conclusions:

• Machine learning models, particularly the XGBoost-based hybrid approach, provide highly accurate survival predictions for brain metastasis patients post-SRT.
• These predictions enhance clinical decision-making by offering precise insights into expected patient outcomes.
• Integrating such machine learning tools into clinical practice can optimize treatment planning and personalize care for improved patient outcomes.