Jove
Visualize
Contact Us
  1. Home
  3. Research Domains
  5. Information And Computing Sciences
  7. Data Management And Data Science
  9. Query Processing And Optimisation
  11. A 3d Hierarchical Cross-modality Interaction Network Using Transformers And Convolutions For Brain Glioma Segmentation In Mr Images.
  1. Home
  3. Research Domains
  5. Information And Computing Sciences
  7. Data Management And Data Science
  9. Query Processing And Optimisation
  11. A 3d Hierarchical Cross-modality Interaction Network Using Transformers And Convolutions For Brain Glioma Segmentation In Mr Images.

Related Experiment Video

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

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

380

A 3D hierarchical cross-modality interaction network using transformers and convolutions for brain glioma segmentation in MR images.

Yuzhou Zhuang1, Hong Liu1, Wei Fang2

  • 1School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan, China.

Medical Physics
|August 13, 2024

View abstract on PubMed

Summary
This summary is machine-generated.

This study introduces a novel 3D hierarchical cross-modality interaction network (HCMINet) for accurate brain glioma segmentation from multi-parametric MRI scans. The HCMINet effectively segments gliomas, improving diagnosis and reducing radiologist workload.

Area of Science:

  • Medical Imaging
  • Artificial Intelligence
  • Neuroscience

Background:

  • Accurate glioma segmentation from multi-parametric MRI is crucial for diagnosis but challenging due to indistinct tumor boundaries and heterogeneous appearances.
  • Existing 3D segmentation networks struggle with hierarchical interactions between modalities and comprehensive feature representation for all glioma sub-regions.

Purpose of the Study:

  • To develop a 3D hierarchical cross-modality interaction network (HCMINet) for accurate multi-modal glioma segmentation.
  • To leverage hierarchical cross-modality interactions for learning modality-specific and shared knowledge for glioma sub-region segmentation.

Main Methods:

  • Designed a hierarchical cross-modality interaction Transformer (HCMITrans) encoder for hierarchical encoding and fusion of multi-modal features.
Keywords:
MR imagesbrain glioma segmentationcontextual information learningmulti‐modal feature fusionvision transformer

More Related Videos

Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
06:48

Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images

Published on: January 7, 2019

8.9K
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

2.3K

Related Experiment Videos

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

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

380
Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images
06:48

Automated Segmentation of Cortical Grey Matter from T1-Weighted MRI Images

Published on: January 7, 2019

8.9K
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

2.3K

Related Concept Videos

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
  • Employed a dual-encoder architecture combining HCMITrans and a modality-shared convolutional encoder for global and local contextual information learning.
  • Utilized a progressive hybrid context fusion (PHCF) decoder with a local-global context fusion (LGCF) module to integrate features and mitigate contextual discrepancies.

    • Main Results:

    • Evaluated on BraTS2020 and BraTS2021 datasets, demonstrating superior performance over existing Transformer-based and CNN-based methods.
    • Achieved state-of-the-art mean Dice Similarity Coefficient (DSC) values of 85.33% on BraTS2020 and 91.09% on BraTS2021.
    • The proposed HCMINet effectively segments glioma sub-regions from multi-parametric MR images.

    Conclusions:

    • The proposed HCMINet enables accurate and automated glioma segmentation from multi-parametric MR images.
    • This method aids in the quantitative analysis of brain gliomas and reduces the annotation burden for neuroradiologists.