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: May 28, 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

MFDA-UNet: Medical Image Segmentation with Frequency-Decoupled Representation and Gated Cross-Scale Integration.

Weiming Deng1, Cong Wu1

  • 1School of Science, Hubei University of Technology, Wuhan 430068, China.

Sensors (Basel, Switzerland)
|May 27, 2026
PubMed
Summary

MFDA-UNet combines convolution and linear attention for medical image segmentation. This hybrid approach enhances local and global feature extraction, improving segmentation accuracy with efficient computation.

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

Privacy-Enhanced Vertical Federated Learning for Healthcare via Directional Noise and Subset Representations.

IEEE journal of biomedical and health informatics·2026
Same author

Selection of proper artificial intelligence techniques developed for CT scan image analysis of liver cancer using fuzzy AHP-TOPSIS.

BMC medical imaging·2026
Same author

Identification and functional characterization of a novel mutation in the NEUROD1 gene in a Chinese family with maturity-onset diabetes of the young.

Acta diabetologica·2026
Same author

Analysis of Long-Term and Short-Term Efficacy of Different Types of Tympanosclerosis Surgery Under Total Otoendoscopy.

The Annals of otology, rhinology, and laryngology·2026
Same author

MiR-26b-5p Predicts the Severity of Crohn's Disease and the Degree of Malnutrition.

Folia biologica·2026
Same author

Towards precision oncology: unsupervised manifold learning for spatial molecular profiling in cancer tissues.

BMC bioinformatics·2026

Area of Science:

  • Medical Image Analysis
  • Deep Learning Architectures
  • Computer Vision

Background:

  • Convolutional Neural Networks (CNNs) have limited global context perception due to small receptive fields.
  • Transformers offer global context but suffer from quadratic computational costs with increasing image resolution.
  • Existing methods struggle to balance local feature extraction and global context modeling efficiently.

Purpose of the Study:

  • To introduce MFDA-UNet, a novel hybrid deep learning model for medical image segmentation.
  • To address the limitations of CNNs and Transformers in capturing multi-scale semantic information.
  • To achieve efficient and robust medical image segmentation through synergistic feature processing.

Main Methods:

  • Developed Mamba-inspired Frequency-Decoupled Attention (MFDA) blocks for processing high-frequency local and low-frequency global information.
Keywords:
linear attentionmedical image segmentationstate space models

Related Experiment Videos

Last Updated: May 28, 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

  • Introduced Mamba-Enhanced Linear Attention (MELA) blocks incorporating positional encoding and Mamba structures for efficient long-range dependency modeling.
  • Integrated Gated Cross-Scale Attention (GCSA) modules to optimize skip connections via cross-scale attention and gating mechanisms.

    • Main Results:

    • MFDA-UNet demonstrated improved Dice Similarity Coefficient (DSC) scores across four multi-modal benchmarks (ISIC 2017, ISIC 2018, Synapse, ACDC).
    • Achieved performance gains of 0.44%, 0.15%, 0.53%, and 0.84% compared to the second-best models on respective datasets.
    • Effectively captured local and global multi-scale semantics with reduced computational overhead.

    Conclusions:

    • MFDA-UNet offers an efficient and robust solution for medical image segmentation.
    • The hybrid architecture effectively balances local and global feature extraction.
    • The proposed MFDA, MELA, and GCSA modules contribute to enhanced segmentation performance and computational efficiency.