Deep Learning Predicts Colon Capsule Endoscopy Position Across Segments Over Time

Researchers have developed an algorithm that predicts which colon segment a colon capsule endoscopy capsule is in at each time point, addressing a clinical challenge where no method has proven reliable or clinically viable for estimating capsule position as a function of time.

Colon capsule endoscopy serves as a non-invasive investigation method in which a camera capsule travels through and records video of the large bowel. Accurately localizing findings such as polyps during these procedures is important for planning and performing follow-up colonoscopies.

To determine the capsule location, the investigators trained a ResNet50 deep neural network to classify individual endoscopic images as small bowel, ascending colon, transverse colon, or descending colon.

To improve temporal consistency across the examination, the system applies a sliding window approach combined with a matrix of transition weights on top of the per-image classifications. The algorithm provides a position estimate for all time points by filtering out wall-facing images and images of low colon cleaning quality, with the sliding window processing the remaining images.

The study utilized 1,668 videos and clinical reports from the Danish CareForColon2015 trial. Within this dataset, training images were automatically assigned report-derived weak colon segment labels based on landmark timestamps recorded in the trial reports.

On an independent test set extracted from 150 videos, the algorithm achieved a per-image classification accuracy of 0.792. When evaluating polyp colon segment prediction against clinical report annotations across 818 held-out test videos, the approach demonstrated an accuracy of 0.779.

While the methodology can potentially assist clinicians in locating polyps prior to a follow-up colonoscopy, the authors emphasize that the system must still be evaluated in a prospective clinical study.

Subject of Research: Colon capsule endoscopy position estimation

Article Title: Colon capsule endoscopy position estimation based on image classification

Article References: Hindberg, K. D., Sharma, P., Schelde-Olesen, B., Deding, U., Frøkjær Justsen, J. F., Braun, J.-M., & Nadimi, E. S. (2026). Colon capsule endoscopy position estimation based on image classification. BMC Medical Imaging. https://doi.org/10.1186/s12880-026-03023-6

Image Credits: AI Generated

DOI: 10.1186/s12880-026-03023-6

Keywords: colon capsule endoscopy, deep learning, ResNet50, image classification, polyp localization, CareForColon2015, medical imaging, gastroenterology, Colon, capsule, endoscopy, position