Exploratory PET/CT Radiomics for Predicting Early Progression in Locally Advanced Pancreatic Cancer

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Early progression (EP) is a significant challenge in locally advanced pancreatic cancer (LAPC).
• Predicting EP is crucial for effective treatment planning and improving patient outcomes.
• Current prediction methods for EP in LAPC are insufficient.

Purpose of the Study:

• To develop and validate a multiparametric predictive model for early progression in LAPC.
• To integrate radiomic features from ¹⁸F-FDG PET/CT with clinical data.
• To enhance risk stratification for personalized treatment decisions in LAPC.

Main Methods:

• Extraction of 242 radiomic features (first-order, GLCM, LBP-TOP) from CT and PET scans.
• Integration of radiomic features with PET-derived metrics and clinical variables.
• Development of a two-level decision tree classifier with cross-validation and feature selection.

Main Results:

• The multiparametric model achieved 80.7% accuracy and an AUC of 0.83.
• Integrated CT and PET texture analysis identified patients at high risk of EP.
• The model demonstrated effective risk stratification prior to treatment initiation.

Conclusions:

• ¹⁸F-FDG PET/CT radiomic biomarkers combined with clinical data can non-invasively assess tumor heterogeneity.
• This approach improves risk stratification for LAPC patients.
• The findings support personalized therapeutic decision-making for LAPC.