Deep learning model predicts organ failure in acute pancreatitis hours in advance

Acute pancreatitis (AP) is one of the most common acute digestive disorders worldwide, and its most feared complication—persistent organ failure—drives mortality rates of 30% to 50%. Yet current scoring systems and computed tomography (CT)-based assessments often miss early signs of organ failure or depend on delayed laboratory results, limiting their ability to guide timely treatment.

A new study published online in the Journal of Pancreatology on August 14, 2026, reports an artificial intelligence (AI) tool that addresses this gap. Researchers at Changhai Hospital and Shanghai 411 Hospital developed ORACLE (Organ failure Risk Assessment with CT and Learning Engine), which combines deep learning radiomics extracted from multiphase CT scans with clinical variables to predict organ failure automatically.

The multicenter study enrolled 2,746 patients with AP (2011–2024), split into training, validation, and independent external test cohorts. The ORACLE model achieved areas under the curve (AUCs) of 0.85, 0.89, and 0.81 across these cohorts—significantly outperforming the Modified CT Severity Index (AUC 0.68–0.74) and clinical models (AUC 0.67–0.71).

Notably, the model delivered a median early warning of 3.5 hours before organ failure became clinically apparent, with 55% of cases predicted at least 3 hours in advance. Among high-risk patients flagged by the model, the incidence of organ failure reached 92.1%. The overall negative predictive value was 97.2%, meaning a low-risk result reliably ruled out organ failure.

By converting routine CT images into an automated early-warning system for organ failure, the model offers a practical tool for earlier identification of high-risk patients and more targeted allocation of intensive monitoring resources.

The authors note that the tool’s generalizability was confirmed through independent multicenter validation. They envision that automated, imaging-based risk stratification could become a practical adjunct in emergency and critical care settings, enhancing early intervention strategies and optimizing resource allocation.

***


Reference
Title of original paper: Prediction of organ failure in acute pancreatitis via CT: A multicenter deep learning model with early clinical utility
Journal: Journal of Pancreatology
DOI: http://doi.org/10.1097/JP9.0000000000000269

About Yun Bian from Changhai Hospital, China
Yun Bian
Department of Radiology,
Changhai Hospital, Shanghai 200433, China.
E-mail: bianyun2012@foxmail.com

Funding information
This research was supported in part by National Science Foundation for Scientists of China (Nos. 81871352, 82171915, 82171930, 82202214, 82271972, 82371955, 82202125), Natural Science Foundation of Shanghai Science and Technology Innovation Action Plan (Nos. 21ZR1478500, 21Y11910300), Clinical Research Plan of Clinical Research Plan of Shanghai Hospital Development Center (SHDC) (No. SHDC2022CRD028), Shanghai Municipal Health Commission (No. 2024ZZ1015), and Plan for Promoting Scientific Research Paradigm Reform through AI (No. 2024RGZD001).