Machine learning for the rElapse risk eValuation in acute biliary pancreatitis: The deep learning MINERVA study...

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Gastroenterology

Background:

• Mild acute biliary pancreatitis (MABP) poses significant clinical and economic burdens due to potential relapse.
• Current guidelines recommend early cholecystectomy (EC) to prevent recurrent acute pancreatitis (RAP), but practice varies.
• There is a need for reliable tools to predict RAP risk in MABP patients.

Purpose of the Study:

• To develop and validate a machine learning (ML) model to predict the risk of RAP in MABP patients.
• To enhance clinical decision-making for MABP management.
• To provide a reliable, cost-effective, and accessible predictive tool for healthcare professionals.

Main Methods:

• The MINERVA study will collect retrospective and prospective data from adult MABP patients not undergoing EC during index admission.
• A machine learning model using convolutional neural networks (CNN) will be developed for risk prediction.
• The model involves data transformation (kPCA), image creation, convolutional filtering, and risk assessment using performance metrics like AUC.

Main Results:

• The study aims to validate the ML model's performance in predicting RAP at 30, 60, 90 days, and 1-year.
• Performance metrics including accuracy, precision, recall, and AUC will be used for evaluation.
• The developed MINERVA score is expected to offer reliable risk stratification.

Conclusions:

• The MINERVA study addresses the gap in predicting RAP risk in MABP patients using advanced ML.
• The practical application of AI in clinical settings can potentially reduce RAP incidence and healthcare costs.
• The MINERVA score aims to be a valuable tool for optimizing MABP patient care.