Predicting distant failure in early stage NSCLC treated with SBRT using clinical parameters

Area of Science:

• Oncology
• Radiation Oncology
• Medical Informatics
• Machine Learning in Medicine

Background:

• Early-stage non-small cell lung cancer (NSCLC) requires effective treatment strategies.
• Stereotactic body radiation therapy (SBRT) is a key treatment for early NSCLC.
• Predicting distant failure is crucial for optimizing patient outcomes after SBRT.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting early distant failure in NSCLC patients.
• To identify optimal clinical parameters for enhancing predictive accuracy.
• To compare the performance of different machine learning algorithms and parameter selection strategies.

Main Methods:

• Utilized a dataset of 81 early-stage NSCLC patients treated with SBRT.
• Developed predictive models using Artificial Neural Network (ANN), Logistic Regression (LR), and Support Vector Machine (SVM).
• Employed Clonal Selection Algorithm (CSA), Sequential Forward Selection (SFS), and Statistical Analysis (SA) for parameter selection.
• Validated models using 5-fold cross-validation and assessed performance via AUC, sensitivity, and specificity.

Main Results:

• Support Vector Machine (SVM) achieved the highest Area Under the Curve (AUC) of 0.80.
• SVM demonstrated superior sensitivity (83.1%) and specificity (63.6%) compared to ANN and LR.
• The Clonal Selection Algorithm (CSA) based parameter selection strategy outperformed SFS and SA.

Conclusions:

• The SVM model, combined with the CSA parameter selection strategy, effectively predicts distant failure in NSCLC patients treated with SBRT.
• Clinical parameters can be leveraged by machine learning to improve treatment planning and patient monitoring.
• This approach offers a promising tool for personalized medicine in radiation oncology.