A Nomogram for Early Prediction of Infected Pancreatic Necrosis Based on Contrast-Enhanced CT Radiomics and...

Area of Science:

• Radiology
• Medical Imaging
• Machine Learning

Background:

• Acute necrotizing pancreatitis (ANP) is a severe condition.
• Early and accurate identification of infected pancreatic necrosis (IPN) is crucial for effective management.
• Current diagnostic methods may have limitations in early detection.

Purpose of the Study:

• To develop and validate a nomogram for early and accurate prediction of IPN in ANP patients.
• To integrate clinical data with radiomic features from contrast-enhanced computed tomography (CECT).

Main Methods:

• Retrospective study of 203 ANP patients.
• Radiomic features extracted from CECT images using PyRadiomics.
• Machine learning classifiers (LightGBM) used to build a radiomic signature.
• Multivariate logistic regression to develop a combined clinical-radiomic nomogram.

Main Results:

• The combined nomogram demonstrated high predictive efficacy (AUC 0.877 training, 0.829 testing).
• APACHE II and MCTSI scores were identified as independent predictors.
• The model showed good calibration and significant clinical utility via decision curve analysis.

Conclusions:

• The developed clinical-radiomic nomogram is effective for predicting IPN in ANP patients.
• This tool offers a quantitative, non-invasive approach for early IPN detection.
• Potential to improve patient outcomes through timely intervention.