Delta Radiomic Features Predict Resection Margin Status and Overall Survival in Neoadjuvant-Treated Pancreatic Cancer...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Neoadjuvant therapy (NAT) is standard for pancreatic ductal adenocarcinoma (PDAC) surgery.
• Predicting resection margin status (RMS) and prognosis preoperatively is essential due to surgery's morbidity.
• Radiomic models, specifically delta radiomic features (DRFs), offer potential for improved preoperative predictions by analyzing treatment dynamics.

Purpose of the Study:

• To develop and validate predictive models for RMS, overall survival (OS), and disease-free survival (DFS) in PDAC patients undergoing NAT.
• To assess the performance of models combining clinical variables and DRFs compared to clinical or radiomics-only models.

Main Methods:

• Retrospective collection of clinical, pathological, and CT scan data from PDAC patients receiving NAT.
• Extraction of 257 radiomics features and calculation of DRFs by subtracting pre- and post-NAT features.
• Construction of Cox proportional and binary prediction models incorporating clinical variables and DRFs to predict OS, DFS, and RMS.

Main Results:

• A combined model integrating clinical variables (pre-NAT CA19-9, ∆CA19-9, resectability) and DRFs outperformed clinical or radiomics-only models.
• The combined model demonstrated strong predictive performance for OS, DFS, and RMS in both discovery and validation cohorts.
• External validation confirmed the model's robustness across different institutions and NAT regimens.

Conclusions:

• Externally validated delta-radiomics models incorporating clinical variables show significant promise.
• These models can aid in predicting RMS, OS, and DFS for NAT-treated PDAC patients.
• This approach may enhance preoperative risk stratification and treatment planning for PDAC.