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

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

AnyMetric: Universal Metric Optimization for Medical Image Segmentation With Preference Learning.

Chaoqun Wang, Jiawei Mo, Shaobo Min

    IEEE Journal of Biomedical and Health Informatics
    |June 10, 2026
    PubMed
    Summary

    Preference-Guided Segmentation Network (PGSNet) optimizes medical image segmentation by learning human preferences, avoiding complex surrogate loss functions. This approach achieves superior segmentation results across various metrics and datasets.

    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

    LTF promotes central nervous system leukemia progression via neutrophil serine proteases.

    Frontiers in pharmacology·2026
    Same author

    Changes in ovarian reserve function after laparoscopic ovarian cystectomy: a retrospective cohort study.

    Frontiers in surgery·2026
    Same author

    Comparative efficacy of double and triple arthrodesis in adult-acquired flatfoot deformity: a retrospective analysis.

    BMC surgery·2026
    Same author

    High Tree Species Diversity Promotes Thermal Enhancement Response of Microbial Carbon Use Efficiency.

    Advanced science (Weinheim, Baden-Wurttemberg, Germany)·2026
    Same author

    A Hydrogel Delivery System Based on Selenium Nanoparticles and bFGF for Promoting the Repair of Skin Wounds.

    Biomedicines·2026
    Same author

    Sex and social group altered the gut microbiome and fecal metabolome in the critically endangered Yangtze finless porpoise (<i>Neophocaena asiaeorientalis asiaeorientalis</i>).

    Current zoology·2026

    Area of Science:

    • Medical Imaging
    • Computer Vision
    • Machine Learning

    Background:

    • Accurate medical image segmentation is vital for disease diagnosis and treatment planning.
    • Current methods rely on complex, handcrafted surrogate loss functions for optimizing segmentation metrics.
    • Customizing loss functions for diverse and non-differentiable metrics is challenging and labor-intensive.

    Purpose of the Study:

    • To introduce a novel framework, Preference-Guided Segmentation Network (PGSNet), for unified optimization of arbitrary medical image segmentation metrics.
    • To eliminate the need for handcrafted surrogate loss functions by directly learning evaluation preferences.
    • To provide a generalizable and flexible optimization framework for various segmentation metrics.

    Main Methods:

    • PGSNet generates diverse segmentation candidates using multi-scale representations.

    Related Experiment Videos

    Last Updated: Jun 12, 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

  • A reward function, derived from target evaluation metrics, ranks these candidates based on segmentation preferences.
  • Preference signals guide model training to favor superior segmentation outcomes over inferior ones.

    • Main Results:

    • PGSNet achieves superior segmentation results compared to existing methods on challenging public datasets.
    • The framework demonstrates effective optimization across various arbitrary evaluation metrics.
    • Experiments confirm the generalizability and flexibility of the proposed preference-based optimization strategy.

    Conclusions:

    • PGSNet offers a novel and effective approach to medical image segmentation optimization.
    • The framework successfully bypasses the limitations of handcrafted surrogate loss functions.
    • PGSNet provides a flexible and generalizable solution for optimizing diverse segmentation evaluation metrics.