MoLPre: A Machine Learning Model to Predict Metastasis of cT1 Solid Lung Cancer

Area of Science:

• Oncology
• Medical Informatics
• Machine Learning

Background:

• Over 20% of patients with clinical T1 (cT1) solid non-small cell lung cancer (NSCLC) present with nodal or extrathoracic metastasis.
• Early detection of metastasis is critical for optimizing therapeutic strategies and patient risk stratification in clinical practice.

Purpose of the Study:

• To develop and validate a predictive model for metastasis in patients with cT1 solid NSCLC.
• To identify key clinical features influencing metastasis prediction.
• To create a user-friendly tool for clinical application.

Main Methods:

• Utilized clinicopathological data from 2018-2022 for early-stage solitary lung cancer patients.
• Employed Random Forest and Shapley Additive Explanations (SHAP) for feature selection, identifying 9 key predictors.
• Developed and compared Random Forest, Gradient Boosting, and AdaBoost classifiers for metastasis prediction.

Main Results:

• The Random Forest model achieved an average precision of 0.93 and an Area Under the Curve (AUC) of 0.92 (95% CI: 0.88-0.94).
• Outperformed Gradient Boosting (AUC 0.87) and AdaBoost (AUC 0.90) classifiers in internal validation.
• The model demonstrated superior performance across five diagnostic indices.

Conclusions:

• The developed Random Forest model accurately predicts metastasis in cT1 solid NSCLC.
• The MoLPre web application (https://molpre.cqmu.edu.cn/) integrates this model for practical clinical decision support.
• This tool can enhance metastasis prediction, aiding therapeutic planning and risk stratification.