PREDICTION OF POSTOPERATIVE PANCREATIC FISTULAS BY RADIOMIC ANALYSIS OF PREOPERATIVE CT SCANS
Uwe A. Wittel*, Johannes D. Lettner, Jakob Neubauer
Albert-Ludwigs-Universitat Freiburg Medizinische Fakultat, Freiburg, Baden-Württemberg, Germany
Objective: This study employs radiomics to determine whether postoperative pancreatic fistula (POPF) can be predicted prior to a pancreaticoduodenectomy with a POPF-prediction-model (PPM). Subsequently, the results were compared with Roberts' preoperative score. In addition, we investigated the prediction of perioperative amylase dynamics by radiomics with an amylase-prediction-model (APM) to translate radiological findings into clinical relevance.
Methods: 68 patients met the inclusion requirements. Radiomics features were extracted from the arterial phase of computed tomography (CT) at 1mm slice thickness of each patient. Afterwards the top 5% of the most significant features were correlated with the development of POPF. For APM, the radiomics features were correlated with postoperative maximum drain-amylase levels at a cut-off by 1000U/l.
Results: PPM predicted POPF with an AUC of 0.897 (CI=82.3-97.1) in the cohort. The AUC of PPM was higher than Roberts' score, but without statistical significance. The attempt to predict postoperative amylase dynamics in the drainage fluid at a cut-off of 1000U/l was also successful. APM achieved an AUC of 0.936 (CI=88%-99.1%).
Conclusion: It was shown that preoperative prediction of POPF by radiomics was successful. Likewise, APM showed promising results regarding postoperative amylase progression and drainage handling. Thus, both models offer new approaches in the management of POPF.
Back to 2023 Abstracts