Predictive models for intestinal obstruction: from clinical scores to artificial intelligence

Area of Science:

• Surgical Emergency Research
• Medical Informatics
• Clinical Decision Support Systems

Background:

• Intestinal obstruction is a critical surgical emergency where delayed recognition of complications like strangulation or ischemia can be fatal.
• Existing prediction models aim to diagnose obstruction, predict urgent surgery, identify strangulation/ischemia, and forecast treatment failure.

Purpose of the Study:

• To critically review existing prediction models for intestinal obstruction.
• To clarify the clinical scope, validation status, and translational limitations of these models.

Main Methods:

• A comprehensive PubMed search was conducted for studies on prediction models for intestinal obstruction, including AI and CT-based approaches.
• Eligible articles were synthesized narratively due to heterogeneity in study design, endpoints, and modalities.

Main Results:

• Conventional scores are accessible but variable; CT-integrated models enhance risk assessment but require expertise.
• Machine learning and deep learning models show promise but often lack external validation and standardization.
• Direct comparison of model performance (e.g., AUC values) across studies is cautioned due to heterogeneity.

Conclusions:

• The field is evolving towards dynamic, multimodal decision-support systems.
• Key research priorities include prospective multicenter validation, standardized definitions, explainability, and workflow integration.