Deep Learning in Classifying Bowel Obstruction Radiographs

Self-supervised Learning for Classifying Bowel Obstruction on Upright Abdominal Radiography

Background: Accurate labeling of obstruction site on upright abdominal radiograph is a challenging task. The lack of ground truth leads to poor performance on supervised learning models. To address this issue, self-supervised learning (SSL) is proposed to classify normal, small bowel obstruction (SBO), and large bowel obstruction (LBO) radiographs using a few confirmed samples.

Methods: A few number of confirmed and a large number of unlabeled radiographs were categorized based on the ground truth. The SSL model was firstly trained on the unlabeled radiographs, and then fine-tuned on the confirmed radiographs. ResNet50 and VGG16 were used for the embedded base encoders, whose weights and parameters were adjusted during training process. Furthermore, it was tested on an independent dataset, compared with supervised learning models and human interpreters. Finally, the t-SNE and Grad-CAM were used to visualize the model's interpretation.

Study Overview

• Status: Active, not recruiting
• Conditions: Polyp of Colon; Bowel Disease; Digestive System Disease

• Study Type: Observational
• Enrollment (Estimated): 4500

Contacts and Locations

Study Locations

• China
  • Jiangsu
    • Suzhou, Jiangsu, China, 215006
      • TheFirst Affiliated Hospital of Soochow University

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: Adult; Older Adult
• Accepts Healthy Volunteers: No

• Sampling Method: Probability Sample

Study Population

participants received upright abdominal radiographs were included in this study. They can be divided with participants with normal abdominal radiographs, participants with small bowel obstruction, and participants with large bowel obstruction. The study strictly follow the inclusion and exclusion criteria.

Description

Inclusion Criteria:

1. The hospital imaging system looked for plain abdominal standing films diagnosed as intestinal obstruction or normal between 2022 and 2024
2. Aged 18 to 80 years
3. The main complaint was gastrointestinal symptoms

Exclusion Criteria:

1. Image interference, fuzzy performance, difficult to distinguish
2. Non-gastrointestinal symptoms were the main complaint
3. Supine, prone, or lateral decubitus radiography
4. Paralytic obstruction, closed loop obstruction, et al

Study Plan

How is the study designed?

Number of groups / cohorts

3

Cohorts and Interventions

Group / Cohort
patients with normal abdominal radiographs; patients with normal abdominal radiographs, which were confirmed by extra imaging examinations and clinical data. The imaging examinations comprised CT, magnetic resonance imaging (MRI) and colonoscopy in the subsequent 72 hours, while clinical data included recent hospital admission information and surgical operation notes.
patients with small bowel obstruction radiographs; patients with small bowel obstruction radiographs, which were confirmed by extra imaging examinations and clinical data. The imaging examinations comprised CT, magnetic resonance imaging (MRI) and colonoscopy in the subsequent 72 hours, while clinical data included recent hospital admission information and surgical operation notes. In terms of location, small-bowel obstruction (SBO) involves the duodenum, jejunum, and ileum
patients with large bowel obstruction radiographs; patients with large bowel obstruction radiographs, which were confirmed by extra imaging examinations and clinical data. The imaging examinations comprised CT, magnetic resonance imaging (MRI) and colonoscopy in the subsequent 72 hours, while clinical data included recent hospital admission information and surgical operation notes. In terms of location, large-bowel obstruction (SBO), involves the cecum, colon, and rectum.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Diagnostic and classification performance	Accuracy, Recall, Precision, F1-score and confusion matrix	1 week

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Visualized interpretation of the self-supervised model	Grad-CAM and t-SNE to visualize the interpretation of the SSL model	1 week

Collaborators and Investigators

• Sponsor: The First Affiliated Hospital of Soochow University
• Investigators: Study Director: Rui Li, MD, The First Affiliated Hospital of Soochow University

Publications and helpful links

General Publications

• Markogiannakis H, Messaris E, Dardamanis D, Pararas N, Tzertzemelis D, Giannopoulos P, Larentzakis A, Lagoudianakis E, Manouras A, Bramis I. Acute mechanical bowel obstruction: clinical presentation, etiology, management and outcome. World J Gastroenterol. 2007 Jan 21;13(3):432-7. doi: 10.3748/wjg.v13.i3.432.
• Cheng PM, Tran KN, Whang G, Tejura TK. Refining Convolutional Neural Network Detection of Small-Bowel Obstruction in Conventional Radiography. AJR Am J Roentgenol. 2019 Feb;212(2):342-350. doi: 10.2214/AJR.18.20362. Epub 2018 Nov 26.
• Kim DH, Wit H, Thurston M, Long M, Maskell GF, Strugnell MJ, Shetty D, Smith IM, Hollings NP. An artificial intelligence deep learning model for identification of small bowel obstruction on plain abdominal radiographs. Br J Radiol. 2021 Jun 1;94(1122):20201407. doi: 10.1259/bjr.20201407. Epub 2021 Apr 27.
• Frager D. Intestinal obstruction role of CT. Gastroenterol Clin North Am. 2002 Sep;31(3):777-99. doi: 10.1016/s0889-8553(02)00026-2.
• Cappell MS, Batke M. Mechanical obstruction of the small bowel and colon. Med Clin North Am. 2008 May;92(3):575-97, viii. doi: 10.1016/j.mcna.2008.01.003.
• ten Broek RP, Strik C, Issa Y, Bleichrodt RP, van Goor H. Adhesiolysis-related morbidity in abdominal surgery. Ann Surg. 2013 Jul;258(1):98-106. doi: 10.1097/SLA.0b013e31826f4969.
• Vanderbecq Q, Ardon R, De Reviers A, Ruppli C, Dallongeville A, Boulay-Coletta I, D'Assignies G, Zins M. Adhesion-related small bowel obstruction: deep learning for automatic transition-zone detection by CT. Insights Imaging. 2022 Jan 24;13(1):13. doi: 10.1186/s13244-021-01150-y.
• Chen Y, Mancini M, Zhu X, Akata Z. Semi-Supervised and Unsupervised Deep Visual Learning: A Survey. IEEE Trans Pattern Anal Mach Intell. 2024 Mar;46(3):1327-1347. doi: 10.1109/TPAMI.2022.3201576. Epub 2024 Feb 6.
• Li G, Togo R, Ogawa T, Haseyama M. Self-supervised learning for gastritis detection with gastric X-ray images. Int J Comput Assist Radiol Surg. 2023 Oct;18(10):1841-1848. doi: 10.1007/s11548-023-02891-5. Epub 2023 Apr 11.

Study record dates

Study Major Dates

• Study Start (Actual): December 31, 2022
• Primary Completion (Estimated): April 30, 2024
• Study Completion (Estimated): December 31, 2024

Study Registration Dates

• First Submitted: March 13, 2024
• First Submitted That Met QC Criteria: March 13, 2024
• First Posted (Actual): March 20, 2024

Study Record Updates

• Last Update Posted (Actual): March 20, 2024
• Last Update Submitted That Met QC Criteria: March 13, 2024
• Last Verified: March 1, 2024

More Information

Terms related to this study

• Keywords: Artificial intelligence; Deep learning; Machine learning
• Additional Relevant MeSH Terms: Intestinal Diseases; Pathological Conditions, Anatomical; Polyps; Intestinal Polyps; Gastrointestinal Diseases; Digestive System Diseases; Colonic Polyps; Intestinal Obstruction
• Other Study ID Numbers: 2022098

Plan for Individual participant data (IPD)

• Plan to Share Individual Participant Data (IPD)?: NO

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No