Related Experiment Video
Updated: Oct 1, 2025
Label-free, High-Resolution 3D Imaging and Machine Learning Analysis of Intestinal Organoids via Low-Coherence Holotomography
Published on: August 12, 2025
Untangling and segmenting the small intestine in 3D cine-MRI using deep learning
Louis D van Harten1, Catharina S de Jonge2, Kim J Beek2
1Department of Biomedical Engineering and Physics, Amsterdam University Medical Centers - location AMC, University of Amsterdam, Amsterdam, the Netherlands; Informatics Institute, University of Amsterdam, Amsterdam, the Netherlands.
This study introduces a new 3D cine-MRI analysis method for tracking small intestine movement, improving gastrointestinal disorder diagnosis. The technique enhances visualization and quantitative assessment of intestinal motility.
Area of Science:
- Medical Imaging
- Gastroenterology
- Computational Anatomy
Background:
- Abdominal cine-MRI assesses small intestinal motility, crucial for diagnosing gastrointestinal disorders.
- Current 2D cine-MRI methods limit 3D motility analysis, and deep learning segmentation struggles with intestinal variability.
- A need exists for advanced tools to analyze 3D small intestinal structures and motility.
Purpose of the Study:
- To develop and validate a novel multi-task method for automated tracking and segmentation of the small intestine in 3D cine-MRI.
- To enable detailed, quantitative assessment of small intestinal motility using 3D imaging data.
Main Methods:
- A novel multi-task deep learning approach combining a CNN-based orientation classifier with a stochastic tracker was developed.
- The method performs automated centerline tracking and segmentation of the small intestine in 3D cine-MRI datasets.
- Validation was performed on 3D cine-MRI scans from healthy volunteers and patients with severe bowel disease.
Main Results:
- The method achieved high performance in centerline tracking (F1 score 0.77 in healthy, 0.80 in patients) and segmentation (Dice coefficient 0.88 in healthy, 0.79 in patients).
- Centerline conditioning significantly improved segmentation performance compared to unconditioned networks.
- The approach demonstrated robustness across healthy individuals and patients with severe bowel disease.
Conclusions:
- The developed method provides a robust structural representation of the small intestine from 3D cine-MRI.
- This represents a significant advancement towards automated, quantitative assessment of small intestinal motility.
- The technique holds promise for improved diagnosis and monitoring of gastrointestinal disorders.
