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 12, 2026

Automated Joint Space Detection Improves Bone Segmentation Accuracy
06:45

Automated Joint Space Detection Improves Bone Segmentation Accuracy

Published on: November 28, 2025

Semi-supervised medical image captioning via anatomical collaborative evidence network.

Shengxiang Zhou1, Qiurui Liu2, Liping Cai1

  • 1College of Information Engineering, Sichuan Agricultural University, Ya'an, China.

Frontiers in Medicine
|June 11, 2026
PubMed
Summary
This summary is machine-generated.

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

High rate of chimeric gene origination by retroposition in plant genomes.

The Plant cell·2006
Same author

Enzyme catalysis: tool to make and break amygdalin hydrogelators from renewable resources: a delivery model for hydrophobic drugs.

Journal of the American Chemical Society·2006
Same author

Theoretical probing of deltahedral closo-auroboranes B(x)Au(x)2- (x = 5-12).

Inorganic chemistry·2006
Same author

Density functional theory/time-dependent DFT studies on the structures, trend in DNA-binding affinities, and spectral properties of complexes [Ru(bpy)2(p-R-pip)]2+ (R = -OH, -CH3, -H, -NO2).

The journal of physical chemistry. A·2006
Same author

Sn12(2-): stannaspherene.

Journal of the American Chemical Society·2006
Same author

High efficient mammalian expression and secretion of a functional humanized single-chain Fv/human interleukin-2 molecules.

World journal of gastroenterology·2006

ACE-Net, an Anatomy Collaborative Evidence Network, improves medical image captioning by reducing annotation costs and enhancing anatomical accuracy. This semi-supervised approach uses evidential deep learning for reliable clinical language generation from endoscopic images.

Area of Science:

  • Artificial Intelligence
  • Medical Imaging
  • Computer Vision

Background:

  • Medical image captioning is crucial for clinical language generation but faces challenges with high annotation costs and unreliable image interpretation.
  • Existing methods struggle with hallucinations and overconfidence in ambiguous medical images, particularly in endoscopy.

Purpose of the Study:

  • To develop a semi-supervised medical image captioning model that addresses the limitations of high annotation costs and unreliable visual information.
  • To improve the accuracy and reliability of clinical language generation from medical images, specifically endoscopic visuals.

Main Methods:

  • Proposed ACE-Net (Anatomy Collaborative Evidence Network) integrating evidential deep learning with a soft-gating mechanism to quantify uncertainty and reduce noise.
Keywords:
Mixture-of-Experts (MoE)evidential deep learningimplicit anatomical localizationmedical image captioningsemi-supervised learning

More Related Videos

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

Related Experiment Videos

Last Updated: Jun 12, 2026

Automated Joint Space Detection Improves Bone Segmentation Accuracy
06:45

Automated Joint Space Detection Improves Bone Segmentation Accuracy

Published on: November 28, 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

  • Employed a triple-guided Mixture-of-Experts decoder for organized clinical reasoning (semantic anchoring, visual evidencing, spatial calibration).
  • Utilized a teacher-student co-training framework with spatial consistency alignment for stable anatomical attention without pixel-level supervision.

    • Main Results:

    • ACE-Net achieved a BLEU-4 score of 0.7511 and a ROUGE-L score of 0.8728 on a high-resolution otolaryngology endoscopy dataset.
    • Demonstrated strong text-generation performance and improved anatomical grounding with limited annotations.
    • Validated the effectiveness of evidence-constrained global supervision over expensive pixel-level annotations.

    Conclusions:

    • The proposed ACE-Net offers a data-efficient and reliable paradigm for medical image captioning.
    • Effective anatomical localization can be achieved through global supervision, reducing the need for detailed pixel-level annotations.
    • This approach enhances the practical applicability of AI in generating clinical language from medical images.