66 | wetenschapsdag 2023 Sessie 2b: Classificaties: van Chaos naar Categorie 4 Auteurs E.W. Ingwersen, J.I. Bereska, A. Balduzzi, B.V. Janssen, R. de Robertis, F Struik, C.Y. Nio, J. Stoker, M.G. Besselink, H.A. Marquering, I. Verpalen, F. Daams Abstract titel Radiomics preoperative-Fistula Risk Score (RAD-FRS) for pancreatoduodenectomy: development and external validation Background Accurately predicting the risk of clinically relevant postoperative pancreatic fistula (CR-POPF) after pancreatoduodenectomy prior to surgery may assist surgeons in making more informed treatment decisions and improved patient counseling. The aim was to evaluate the predictive accuracy of a radiomics-based preoperative-Fistula Risk Score (RAD-FRS) for CR-POPF. Methods Radiomic features were derived from preoperative CT scans from adult patients after pancreatoduodenectomy at a single center in the Netherlands (Amsterdam, 2013 -2018) to develop the RADFRS. Extracted radiomic features were analyzed with four machine learning classifiers. The model was externally validated in a single center in Italy (Verona, 2020 – 2021). The RAD-FRS was compared to the Fistula Risk Score (FRS) and the updated alternative Fistula Risk Score (ua-FRS). Results Overall, 359 patients underwent a pancreatoduodenectomy, of whom 89 (25%) developed a CR-POPF. The RAD-FRS model was developed using CT scans of 118 patients, of which three radiomic features where included in the random forest model, and externally validated in 57 patients. The model performed well with an area under the curve (AUC) of 0.90 (95% CI: 0.71 – 0.99) and 0.81 (95% CI: 0.69 – 0.92) in the Amsterdam test set and Verona dataset, respectively. The RAD-FRS performed similarly to the FRS (AUC 0.79) and ua-FRS (AUC 0.79). Conclusion The RAD-FRS, which uses only preoperative CT features, is a new and promising radiomics-based score that has the potential to be integrated with hospital CT report systems and improved patient counseling preoperatively. The model with underlying code is readily available via www.pancreascalculator.com and www.github. com/PHAIR-Consortium/POPF-predictor.
RkJQdWJsaXNoZXIy MTk4NDMw