POPF

POPF predicts the risk of postoperative pancreatic fistula following pancreatoduodenectomy (PD) using artificial intelligence models to support preoperative risk assessment.

Key Features:

• AI-Driven Algorithms: Employs random forest (RF) and neural network (NN) algorithms enhanced with recursive feature elimination (RFE) for model development.
• Comprehensive Data Analysis: Analyzes 38 variables from medical records of 1,769 patients who underwent PD at Samsung Medical Center between 2007 and 2016.
• Risk Factor Identification: Identifies 16 critical risk factors for POPF, including pancreatic duct diameter, body mass index, preoperative serum albumin, lipase, intraoperative fluid infusion volume, age, platelet count, tumor location, combined venous resection, co-existing pancreatitis, neoadjuvant radiotherapy, American Society of Anesthesiologists' score, sex, pancreatic texture, underlying heart disease, and preoperative endoscopic biliary decompression.
• High Predictive Accuracy: Demonstrates an area under the curve (AUC) of 0.74 for the neural network with RFE.

Scientific Applications:

• Preoperative Risk Assessment: Supports preoperative risk assessment and facilitates clinical decision-making for patients undergoing pancreatoduodenectomy.
• Personalized Medicine in Surgical Oncology: Enables nuanced risk stratification to support personalized medicine approaches in surgical oncology.

Methodology:

Computational methods include random forest and neural network models with recursive feature elimination (RFE) applied to 38 clinical variables from 1,769 PD patients, with median imputation used to handle missing data.