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 Experiment Video

Updated: Jun 16, 2026

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

Self-distillation double student network for semi-supervised medical image segmentation.

Huaxiang Liu1, Xin Li1, Jie Jin1

  • 1Taizhou University Affiliated Taizhou Central Hospital, Taizhou University, Taizhou, Zhejiang, China.

Frontiers in Physiology
|June 15, 2026
PubMed
Summary

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

NPC2 Regulates Glioblastoma Cholesterol Metabolism and Malignant Behavior via GPX4-Linked Ferroptosis.

Cell biology international·2026
Same author

Spontaneous Symmetry Breaking via Metal-Triggered Surface Defect Engineering for Durable Piezocatalytic Hydrogen Evolution.

Advanced materials (Deerfield Beach, Fla.)·2026
Same author

Oestrogen synthesis mediated by Hsd17b12 isoforms and dynamic regulation by Cyp19a in the olive flounder (Paralichthys olivaceus).

The FEBS journal·2026
Same author

Van der Waals strain hardening and large uniform tensile elongation in GaSe.

Nature materials·2026
Same author

The Patterns of Altitudinal Gradient Differentiation in the Morphological Traits of <i>Calliptamus italicus</i> (L.) (Orthoptera: Acridoidea) and Their Environmental Driving Mechanisms in the Desert Steppe in the Ili River Basin.

Insects·2026
Same author

Plastic-Deformation-Free Fabrication of Freestanding Metal Chalcogenide Ribbons for High-Performance Flexible Thermoelectric Generators.

Advanced materials (Deerfield Beach, Fla.)·2026

This study introduces SDBS-Net, a new semi-supervised learning method for 3D medical image segmentation. It effectively uses unlabeled data to improve segmentation accuracy, especially at boundaries, outperforming existing techniques.

Area of Science:

  • Medical Imaging
  • Artificial Intelligence
  • Computer Vision

Background:

  • Semi-supervised learning (SSL) reduces annotation costs in medical image segmentation by using unlabeled data.
  • Existing Mean-Teacher models have limitations in bidirectional knowledge transfer and handling low-confidence predictions, impacting boundary accuracy.

Purpose of the Study:

  • To propose SDBS-Net, a novel semi-supervised dual-student self-distillation network for 3D medical image segmentation.
  • To enhance segmentation accuracy and boundary delineation by enabling effective bidirectional knowledge transfer between dual student networks.

Main Methods:

  • Developed SDBS-Net with a shared encoder and two parallel decoders (Model I and Model II) for simultaneous processing of labeled and unlabeled data.
  • Introduced Prior Knowledge Learning (PKL) with Hierarchical Difference (HD) modules for multi-scale feature injection.
Keywords:
dual-student frameworkhierarchical feature differencemedical image segmentationself-distillationsemi-supervised learning

Related Experiment Videos

Last Updated: Jun 16, 2026

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

  • Implemented Self-Distillation Learning (SDL) with multi-level soft-label and feature distillation losses for mutual refinement.

    • Main Results:

    • SDBS-Net achieved high Dice scores on the Left Atrium (LA) and Pancreas-CT datasets with limited labeled data (10% and 20%).
    • On the LA dataset, Dice scores reached 88.19% (10% labeled) and 90.51% (20% labeled).
    • On the Pancreas-CT dataset, Dice scores were 71.39% (10% labeled) and 79.37% (20% labeled), outperforming state-of-the-art methods.

    Conclusions:

    • SDBS-Net demonstrates superior performance in 3D medical image segmentation compared to existing semi-supervised approaches.
    • The proposed PKL and SDL schemas effectively improve discriminative representation, boundary delineation, and handle low-confidence predictions.
    • The method significantly reduces reliance on extensive annotations while achieving near fully supervised performance.