Related Experiment Video
Updated: May 20, 2026
07:33
A Mouse Model of Intestinal Partial Obstruction
Published on: March 5, 2018
Development and validation of an explainable machine learning model for predicting surgical intervention in pediatric
Zhendi Tang1,2, Yonghua Fu1,2, Jiajia Zhou1,2
1Department of General Surgery and Trauma Surgery, Children's Hospital of Chongqing Medical University, Chongqing, China.
European Journal of Pediatrics
|May 18, 2026
Summary
This study developed an explainable machine learning model to predict surgical needs in pediatric intestinal obstruction, a condition risking intestinal necrosis. The model provides transparent decision support for clinicians, improving diagnostic accuracy and patient care.
Area of Science:
- Pediatric Surgery
- Medical Artificial Intelligence
- Clinical Decision Support Systems
Background:
- Pediatric intestinal obstruction is a critical condition with a risk of intestinal necrosis.
- Current surgical decision-making lacks standardized criteria, leading to diagnostic variability.
Purpose of the Study:
- To develop an explainable machine learning (ML) model for predicting surgical indication in pediatric intestinal obstruction.
- To reduce diagnostic variability and provide transparent decision support to clinicians.
Main Methods:
- Retrospective study of 642 children (training/internal validation) and 137 (external validation).
- Developed and evaluated 11 ML algorithms using 35 features.
- Utilized SHapley Additive exPlanations (SHAP) for model interpretability.
- Defined surgical indication based on clinical signs, radiographic evidence, or failed conservative management.
Main Results:
- The random forest model achieved high discriminative performance.
- At the optimal threshold, the model demonstrated 89% sensitivity and 99% specificity (internal validation).
- An interpretable 8-feature model achieved AUCs of 0.988 (internal) and 0.981 (external validation).
Conclusions:
- An explainable ML model effectively predicts surgical need in pediatric intestinal obstruction.
- The model offers transparent, interpretable decision support, mitigating the 'black-box' concern.
- This tool is associated with predicting intestinal necrosis and can aid clinical decision-making.
