Classification of Infected Necrotizing Pancreatitis for Surgery Within or Beyond 4 Weeks Using Machine Learning

Area of Science:

• Surgical outcomes
• Machine learning in medicine
• Pancreatic diseases

Background:

• Surgical timing for necrotizing pancreatitis is debated, lacking randomized controlled trial (RCT) resolution.
• Infected pancreatic necrosis presents significant clinical challenges.
• Optimal surgical intervention timing remains a critical unresolved issue.

Purpose of the Study:

• To classify surgical timing (within or beyond 4 weeks) for infected necrotizing pancreatitis using machine learning.
• To identify key predictors for surgical timing and postoperative mortality.
• To leverage advanced computational methods for clinical decision support.

Main Methods:

• Analysis of 223 patients with infected pancreatic necrosis.
• Application of logistic regression, support vector machine, and random forest models.
• Inclusion of generative adversarial networks (GANs) for enhanced classification.

Main Results:

• Interleukin-6, infected necrosis, fever onset, and C-reactive protein identified as key factors for surgical timing.
• Modified Marshall score (admission and preoperational) linked to early surgery mortality.
• Preoperational modified Marshall score, surgery time, organ failure duration, and renal failure onset predict delayed surgery mortality.

Conclusions:

• Machine learning models demonstrate efficacy in predicting surgical intervention timing for infected necrotizing pancreatitis.
• Identification of critical factors influencing both surgical timing and postoperative survival.
• Provides valuable insights for optimizing treatment strategies in complex pancreatic cases.