¹⁸F-FDG PET/CT-based Radiomics Analysis of Different Machine Learning Models for Predicting Pathological Highly...

Area of Science:

• Oncology
• Radiology
• Artificial Intelligence

Background:

• Accurate prediction of pathological invasiveness is crucial for early-stage non-small cell lung cancer (NSCLC).
• Integrating clinicoradiological and radiomic data can enhance diagnostic accuracy.

Purpose of the Study:

• Develop and validate machine learning models to predict pathological high invasiveness in cT1-sized NSCLC.
• Utilize 2-[18F]-fluoro-2-deoxy-D-glucose (18F-FDG) PET/CT imaging features.

Main Methods:

• Retrospective analysis of 1459 NSCLC patients.
• Extraction of 1145 radiomic features from PET/CT scans.
• Development of a combined model using logistic regression, random forest, SVM, and XGBoost algorithms.

Main Results:

• The radiomics model achieved AUCs up to 0.859.
• The XGBoost combined model showed the best predictive performance with AUCs up to 0.958.
• The XGBoost model demonstrated good calibration and high net benefit.

Conclusions:

• The XGBoost combined model is effective for predicting pathological high invasiveness in cT1-sized NSCLC.
• This approach offers a promising tool for improving clinical decision-making in NSCLC management.