Machine-learning-model based tool for screening bone metastases from lung cancer patients in primary care practice

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Bone metastasis (BM) is a significant complication in lung cancer patients.
• Accurate prediction of BM is crucial for timely intervention and improved patient outcomes.

Purpose of the Study:

• To develop and validate machine learning models for predicting bone metastasis in lung cancer patients.
• To identify key hematological indicators associated with bone metastasis.
• To create a user-friendly online tool for real-time BM risk assessment.

Main Methods:

• Utilized a dataset of 8,612 lung cancer patients, including baseline characteristics and hematological data.
• Employed machine learning techniques, including LASSO for variable selection and nine distinct models for prediction.
• Developed and validated models using training and validation sets (7:3 ratio) with under-sampling for imbalance.
• Identified the eXtreme Gradient Boosting (XGBoost) model as the optimal performer.

Main Results:

• Identified nine key indicators for BM prediction: ALP, LYM%, HCT, FBG, TT, TBIL, smoking status, DBIL, and D-dimer.
• The XGBoost model demonstrated strong predictive performance with an AUC of 0.8423 in the training set and 0.7871 in the validation set.
• An online tool was developed based on the XGBoost model for real-time BM prediction in lung cancer patients.

Conclusions:

• A robust machine learning model, XGBoost, effectively predicts bone metastasis in lung cancer patients using hematological indicators.
• The developed online tool provides a valuable resource for clinicians to guide diagnostic evaluations and interventions for high-risk patients.
• This approach facilitates early detection and management of bone metastasis, potentially improving patient prognosis.