Interpretable deep learning model for pediatric strangulated small bowel obstruction on CT: A multicenter study

Clinical Snapshot

Key Findings

• P Value: p = 0.01 for both MIL and combined models in external test cohort

• Effect Size: MIL model AUC 0.86 in external test cohort; combined (MIL + clinical) model AUC 0.87 in external test cohort

• Primary Outcome: Discrimination of strangulated small bowel obstruction (StSBO) from simple small bowel obstruction (SiSBO) in paediatric patients using CT-based MIL model

• Nnt Or Sensitivity: Sensitivity and specificity not reported in abstract. AUC increases with MIL assistance: junior radiologists +16%, experienced radiologists +2%, surgeons +20%. Likelihood ratios cannot be calculated from available data.

• Confidence Interval: MIL model: 95% CI 0.70–1.00; Combined model: 95% CI 0.72–1.00

Clinical Application

Implementation would require integration of the MIL model into existing PACS/radiology workflows, standardised CT acquisition protocols compatible with model training data, and prospective validation in the target clinical environment. The model's dependence on specific CT parameters and institutional data characteristics may limit plug-and-play deployment. Computational infrastructure requirements are not described. In Australia, paediatric SBO is managed at specialist children's hospitals (e.g., Royal Children's Hospital Melbourne, Sydney Children's Hospital, Women's and Children's Hospital Adelaide) under paediatric surgical and radiology teams. CT use in children is subject to ALARA radiation principles and RANZCR guidance. No equivalent deep learning tool is currently TGA-registered or endorsed by RACS or RANZCR for this indication. The RACGP and RACP do not have specific guidelines for paediatric SBO imaging decision support. Any clinical deployment would require TGA Software as a Medical Device (SaMD) regulatory approval under the Therapeutic Goods Act 1989. The PBS does not currently fund AI-assisted diagnostic tools for this indication. Australian paediatric SBO epidemiology (including adhesional vs. other aetiologies) may differ from the Chinese cohort studied. Paediatric patients aged 1–14 years presenting to hospital with CT-confirmed small bowel obstruction requiring differentiation of strangulated from simple obstruction to guide urgency of surgical intervention

Abstract

OBJECTIVE: To develop and validate a deep learning-based multi-instance learning (MIL) model that integrates CT imaging and clinical data to improve the accuracy of discriminating between strangulated small bowel obstruction (StSBO) from simple small bowel obstruction (SiSBO) in pediatric patients. MATERIALS AND METHODS: This multicenter retrospective study, conducted between January 2018 and June 2024, enrolled hospitalized pediatric patients aged 1-14 years with a diagnosis of small bowel obstruction. We developed the clinical, multi-instance learning, and combined models based on CT and clinical features. Model performance was evaluated using receiver operating characteristic (ROC) analysis, while SHapley Additive exPlanations (SHAP) interpreted feature contributions. We further assessed whether MIL-assisted diagnosis could enhance physician accuracy in diagnosing StSBO. RESULTS: The study sample comprised 168 patients (mean age, 6.36 ± 3.97, 118 men). Ascites and closed-loop sign were identified as independent predictors of StSBO on multivariate analysis (both p< 0.05). The MIL model achieved the area under the curve (AUC) of 0.86 (95%CI 0.70-1.00), p= 0.01 in the external test cohort. The combined model showed the highest diagnostic performance (AUC 0.87, 95%CI 0.72-1.00, p= 0.01) in the external test cohort, with MIL-derived features showing predominant importance in SHAP analysis. Both junior and experienced radiologists and surgeon demonstrated improved diagnostic performance with MIL assistance, showing AUC increases of 16%, 2%, and 20%, respectively. CONCLUSIONS: The MIL model performed well in diagnosing StSBO, and clinical data integration improved its performance. As a decision support tool, the model may aid risk stratification and facilitate timely escalation of care in pediatric StSBO management.

References

1. 1.Chang, N., Liu, X., Liu, P., Gao, H., Lin, N., Chen, X., Cui, L., Jia, H., & Yu, B. (2026). Interpretable deep learning model for pediatric strangulated small bowel obstruction on CT: A multicenter study. Journal of Pediatric Surgery. https://doi.org/10.1016/j.jpedsurg.2026.163075