Machine Learning-Based Prediction of Brain Metastasis at Initial Diagnosis in Small-Cell Lung Cancer: Model...

Area of Science:

• Oncology
• Machine Learning
• Biostatistics

Background:

• Brain metastasis (BM) significantly worsens survival and quality of life for small cell lung cancer (SCLC) patients.
• Early identification of SCLC patients at high risk for BM is critical for timely intervention.

Purpose of the Study:

• To develop and validate a machine learning (ML) model for predicting BM in SCLC patients at diagnosis.
• To create an interpretable and clinically accessible tool for BM risk assessment.

Main Methods:

• Utilized logistic regression to identify BM-associated factors.
• Applied eight ML algorithms, with Extreme Gradient Boosting (XGB) selected as the best performer.
• Employed SHapley Additive exPlanations (SHAP) for model interpretability and developed a web calculator for risk estimation.

Main Results:

• The XGB model demonstrated strong predictive performance with an AUC of 0.8762 and AUPRC of 0.9025 in the validation cohort.
• Key predictors identified by SHAP analysis included age, tumor size, T stage, and presence of bone, lung, or distant lymph node metastasis.
• A web-based calculator was created for individualized BM risk stratification.

Conclusions:

• An interpretable ML model, optimized with XGB, was developed and internally validated for identifying BM in SCLC at diagnosis.
• The web-based tool can assist clinicians in identifying high-risk patients and supplement clinical assessments.
• Prospective validation is recommended prior to routine clinical use.