Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

The CathPilot: Usability and Feasibility Evaluation in Simulated-Use Endovascular Interventions.

Annals of biomedical engineering·2026
Same author

Diversifying Robotic Ileocecectomy Methods: Superior Port Placement in a Patient With an Obstructed Crohn's Disease Flare-Up.

Cureus·2026
Same author

Less Than Full-Time Training in General Surgery: Who Benefits and Who Carries the Cost?

ANZ journal of surgery·2026
Same author

A Presentation of Meckel's Diverticulum With Malignancy.

Cureus·2026
Same author

Ileocecal Duplication Cyst Presenting As Bowel Obstruction in an Adult.

Cureus·2026
Same author

Assessing the Association of Age and Preoperative Sodium Level on Colectomy Outcomes: An NSQIP Study.

The Journal of surgical research·2026

Related Experiment Video

Updated: Sep 3, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.9K

Cardiac MRI segmentation with sparse annotations: Ensembling deep learning uncertainty and shape priors.

Fumin Guo1, Matthew Ng2, Grey Kuling2

  • 1Wuhan National Laboratory for Optoelectronics, Biomedical Engineering, Huazhong University of Science and Technology, Wuhan, China; Department of Medical Biophysics, University of Toronto, Toronto, Canada; Sunnybrook Research Institute, University of Toronto, Toronto, Canada.

Medical Image Analysis
|July 25, 2022
PubMed
Summary

This study introduces novel deep learning methods, Globally Optimal Label Fusion (GOLF) and uncertainty-guided coupled continuous kernel cut (ugCCKC), to improve cardiac MRI segmentation accuracy, even with limited data. These algorithms enhance segmentation quality and enable reliable left ventricular function measurements.

Keywords:
Cardiac MRICoupled continuous kernel cutDeep learning ensembleLabel fusionSegmentationShape priorsUncertainty

More Related Videos

Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
04:25

Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies

Published on: December 15, 2023

2.8K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

488

Related Experiment Videos

Last Updated: Sep 3, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
04:48

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

2.9K
Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies
04:25

Author Spotlight: Bridging Gaps in Anatomy and Establishing a Foundation for Algorithmic Studies

Published on: December 15, 2023

2.8K
Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique
04:48

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

488

Area of Science:

  • Medical Imaging
  • Machine Learning
  • Cardiovascular Imaging

Background:

  • Deep learning performance in cardiac MRI segmentation is often limited by small datasets and sparse annotations.
  • Adapting pre-trained models to new datasets can be challenging due to domain shift.
  • Accurate segmentation is crucial for quantitative analysis of cardiac function.

Purpose of the Study:

  • To develop and evaluate an approach to enhance deep learning-based short-axis cardiac MRI segmentation.
  • To address limitations of small datasets and sparse annotations in training deep learning models.
  • To facilitate broader clinical and research application of deep learning in cardiac MRI.

Main Methods:

  • Developed a Globally Optimal Label Fusion (GOLF) algorithm for consensus segmentation from deep learning ensembles, enforcing spatial smoothness.
  • Employed an uncertainty-guided coupled continuous kernel cut (ugCCKC) algorithm incorporating shape priors (left ventricular myocardium and cavity) and uncertainty measurements.
  • Integrated GOLF and ugCCKC into a deep learning ensemble for refining segmentation on unannotated data and updating models, optimizing ugCCKC via max-flow algorithms.

Main Results:

  • GOLF improved segmentation accuracy (lower average symmetric surface distance) compared to simple averaging methods.
  • ugCCKC with shape priors enhanced Dice similarity coefficient (DSC) and reduced average symmetric surface distance (ASSD).
  • The integrated framework achieved high DSC (0.871-0.893 for LVM, 0.933-0.959 for LVC) on small, sparsely annotated datasets (5-10 subjects, 5-25% slices) and alleviated domain shift.
  • Segmentation-derived left ventricular function measurements showed strong correlations with manual analyses (Pearson r=0.77-1.00).

Conclusions:

  • The developed GOLF and ugCCKC algorithms significantly improve deep learning-based cardiac MRI segmentation, particularly with limited data.
  • The integrated framework enables robust segmentation and accurate functional analysis, even with sparse annotations and domain variability.
  • These approaches hold promise for broader adoption of deep learning in cardiac MRI research and clinical workflows.