Machine learning for optimizing the management of patients with suspected digestive obstruction



Main objectives

The first objective of the project is to build a deep neural network model capable of automated digestive occlusion detection from CT scans.
The following objectives are :
- Build language processing models to obtain labels from radiological reports.
- Classification of colonic versus bowel obstructions
- Classification of functional versus mechanical occlusions
- Radiomics analysis of junction zones
- Automated detection of junction zones
- Improved predictive performance for surgical management

The models obtained will help radiologists and streamline the management of emergency patients.


Digestive obstruction is a frequent reason for emergency consultations. Diagnosis requires an abdomino-pelvic CT scan, and management can be either medical or surgical. It is therefore useful to have a guide to radiological interpretation, from initial triage to assessment of the severity and cause of the obstruction. In addition, it would be useful to develop an aid to management decisions, to avoid delays in surgical management.



Quentin Vanderbecq ()


AP-HP, Sorbonne Center for Artificial Intelligence (SCAI), Inria, Groupe Hospitalier Paris Saint-Joseph

Other projects



US Caractérisation et prédiction de la survenue de formes graves ou létales du COVID-19 à partir des données issues de l’EDS de l’AP-HP

Names of partners involved
AP-HP, Inria & Centrale Supélec

read more