Postoperative Pancreatic Fistula After Pancreatoduodenectomy: Can Radiomics Improve Clinical Risk Scores?

Area of Science:

• Surgical Oncology
• Radiology
• Data Science

Background:

• Radiomics quantifies imaging data for clinical applications, including cancer research.
• Its utility in predicting postoperative complications is an emerging area.

Purpose of the Study:

• To evaluate the added benefit of radiomics to clinical models for predicting clinically relevant postoperative pancreatic fistula (CR-POPF) after pancreatoduodenectomy (PD).

Main Methods:

• A retrospective study analyzed CT scans from patients undergoing PD, extracting radiomic features from the future remnant pancreas.
• Five predictive models were developed using clinical variables and radiomic features: preoperative (PreClin), intraoperative (IntraClin), radiomics alone (Rad), PreClin-Rad, and IntraClin-Rad.
• Models were validated on a test set and a prospective cohort.

Main Results:

• Individual models showed moderate predictive performance (AUC 0.74-0.78).
• Combined models integrating radiomics and clinical data achieved superior AUCs (0.82-0.84).
• The Preoperative Clinical data with Radiomics (PreClin-Rad) model demonstrated the best performance (AUC 0.78) in the prospective validation cohort.

Conclusions:

• Radiomics offers a valuable tool for predicting CR-POPF, outperforming traditional clinical risk scores.
• Combined radiomics and clinical data models provide the most robust prediction of POPF after PD.
• The PreClin-Rad model is highly clinically useful due to its reliance on readily available preoperative data.