Radiomics preoperative-Fistula Risk Score (RAD-FRS) for pancreatoduodenectomy: development and external validation

Area of Science:

• Radiology and Medical Imaging
• Surgical Oncology
• Artificial Intelligence in Medicine

Background:

• Postoperative pancreatic fistula (POPF) is a significant complication after pancreatoduodenectomy.
• Accurate preoperative risk prediction is crucial for surgical decision-making and patient counseling.
• Current risk scores may benefit from enhanced predictive capabilities.

Purpose of the Study:

• To evaluate the predictive accuracy of a novel radiomics-based preoperative-Fistula Risk Score (RAD-FRS) for clinically relevant POPF.
• To compare the performance of RAD-FRS against existing Fistula Risk Score (FRS) and updated alternative FRS.

Main Methods:

• Radiomic features were extracted from preoperative CT scans of patients undergoing pancreatoduodenectomy.
• A random forest model was developed using three radiomic features to create the RAD-FRS.
• The RAD-FRS was internally validated and externally validated in independent cohorts.

Main Results:

• The RAD-FRS demonstrated strong predictive performance with an AUC of 0.90 in the internal test set and 0.81 in the external validation set.
• The RAD-FRS showed comparable performance to the FRS and updated alternative FRS (AUC 0.79 for both).
• The study included 359 patients, with 25% developing clinically relevant POPF.

Conclusions:

• The RAD-FRS is a promising, novel tool for predicting POPF risk using only preoperative CT features.
• Its integration with CT reporting systems could enhance preoperative patient counseling.
• The model and code are publicly available for wider adoption and research.