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: Oct 22, 2025

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

9.1K

Loss Weightings for Improving Imbalanced Brain Structure Segmentation Using Fully Convolutional Networks.

Takaaki Sugino1, Toshihiro Kawase1, Shinya Onogi1

  • 1Department of Biomedical Information, Institute of Biomaterials and Bioengineering, Tokyo Medical and Dental University, Tokyo 101-0062, Japan.

Healthcare (Basel, Switzerland)
|August 27, 2021
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

N-butyl Cyanoacrylate Use in Various Neuroendovascular Diseases.

Journal of neuroendovascular therapy·2025
Same author

Development of automatic landmark identification for mandible using curvature-based registration.

Dento maxillo facial radiology·2025
Same author

<i>RNF213</i> p.Arg4810Lys (c.14429G>A) is associated with extracranial arterial stenosis.

Brain communications·2025
Same author

Outcomes and Complications of Endovascular Treatment in Patients with Moyamoya Disease: A Systematic Review.

World neurosurgery·2025
Same author

Improved segmentation of hepatic vascular networks in ultrasound volumes using 3D U-Net with intensity transformation-based data augmentation.

Medical & biological engineering & computing·2025
Same author

Current molecular understanding of central nervous system schwannomas.

Acta neuropathologica communications·2025
This summary is machine-generated.

Loss weighting strategies improve brain structure segmentation in MR images, addressing class imbalance. Focal and distance map-based weightings significantly boost performance for both binary and multi-class tasks.

Area of Science:

  • Medical Imaging
  • Neuroimaging
  • Computer Vision

Background:

  • Brain structure segmentation from MR images is crucial for clinical applications.
  • Fully convolutional networks are used for automated segmentation but struggle with class imbalance.
  • Class imbalance in medical imaging leads to biased models and poor segmentation of smaller structures.

Purpose of the Study:

  • To investigate the effectiveness of various loss weighting strategies for brain structure segmentation on MR images.
  • To address the class imbalance problem in automated segmentation tasks.
  • To compare different weighting methods with baseline loss functions (cross-entropy and Dice).

Main Methods:

  • Utilized a U-net architecture for segmentation of cerebrum, cerebellum, brainstem, and blood vessels from MR images.
Keywords:
brain structure segmentationclass imbalancefully convolutional networksloss weightingmagnetic resonance images

More Related Videos

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
14:08

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images

Published on: April 13, 2013

43.0K

Related Experiment Videos

Last Updated: Oct 22, 2025

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

9.1K
Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images
14:08

Automated Midline Shift and Intracranial Pressure Estimation based on Brain CT Images

Published on: April 13, 2013

43.0K
  • Implemented and evaluated five loss weighting strategies: inverse frequency, median inverse frequency, focal, distance map-based, and distance penalty term-based weighting.
  • Assessed performance using Dice scores on binary-class and multi-class segmentation tasks.

    • Main Results:

    • The Dice loss function combined with focal weighting achieved a 92.8% average Dice score in binary-class segmentation.
    • Cross-entropy loss functions with distance map-based weighting reached up to a 93.1% Dice score in multi-class segmentation.
    • Both distance map-based and focal weightings demonstrated significant improvements in handling class imbalanced segmentation tasks.

    Conclusions:

    • Loss weighting strategies are effective in mitigating class imbalance for brain structure segmentation on MR images.
    • Focal weighting enhances Dice loss performance, while distance map-based weighting boosts cross-entropy loss performance.
    • These findings offer valuable insights for improving automated segmentation accuracy in clinical neuroimaging.