Adhesion-related small bowel obstruction: deep learning for automatic transition-zone detection by CT

Quentin Vanderbecq,Marc Zins,Axel Dallongeville,Isabelle Boulay-Coletta,Gaspard D’Assignies,Camille Ruppli,Antoine De Reviers,Roberto Ardon

doi:10.1186/s13244-021-01150-y

Abstract

BackgroundTo train a machine-learning model to locate the transition zone (TZ) of adhesion-related small bowel obstruction (SBO) on CT scans.Materials and methodsWe used 562 CTs performed in 2005–2018 in 404 patients with adhesion-related SBO. Annotation of the TZs was performed by experienced radiologists and trained residents using bounding boxes. Preprocessing involved using a pretrained model to extract the abdominopelvic region. We modeled TZ localization as a binary classification problem by splitting the abdominopelvic region into 125 patches. We then trained a neural network model to classify each patch as containing or not containing a TZ. We coupled this with a trained probabilistic estimation of presence of a TZ in each patch. The models were first evaluated by computing the area under the receiver operating characteristics curve (AUROC). Then, to assess the clinical benefit, we measured the proportion of total abdominopelvic volume classified as containing a TZ for several different false-negative rates.ResultsThe probability of containing a TZ was highest for the hypogastric region (56.9%). The coupled classification network and probability mapping produced an AUROC of 0.93. For a 15% proportion of volume classified as containing TZs, the probability of highlighted patches containing a TZ was 92%.ConclusionModeling TZ localization by coupling convolutional neural network classification and probabilistic localization estimation shows the way to a possible automatic TZ detection, a complex radiological task with a major clinical impact.

Highlights

Small bowel obstruction (SBO) is a common nontraumatic surgical emergency, with approximately 400,000 admissions annually in the United States [1]
Modeling transition zone (TZ) localization by coupling convolutional neural network classification and probabilistic localization estimation shows the way to a possible automatic TZ detection, a complex radiological task with a major clinical impact
Transition zone (TZ) location We first analyzed our dataset of 562 annotated computed tomography (CT) scans

Summary

Introduction

Small bowel obstruction (SBO) is a common nontraumatic surgical emergency, with approximately 400,000 admissions annually in the United States [1]. The goal is four-fold: (i) to confirm or refute the diagnosis of SBO and, when SBO is present, (ii) to locate the site of the obstruction, that is, the transition zone (TZ) (iii) to identify the cause, and (iv) to look for complications such as strangulation or perforation. Identifying the TZ or TZs (and determining their number) and establishing their locations is the first step in diagnosing the cause of SBO and differentiating the open-loop and closed-loop mechanisms [5]. A diagnosis of closed-loop SBO independently predicts ischemia [6] and help to decide whether surgery is needed in patients with adhesion-related SBO. To train a machine-learning model to locate the transition zone (TZ) of adhesion-related small bowel obstruction (SBO) on CT scans

Objectives

Methods

Results

Conclusion