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 Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Advances and Challenges in Pharmacokinetic Modeling for PET Imaging: Compartment Models, Input Functions, and Quantitative Techniques.

Tomography (Ann Arbor, Mich.)·2026
Same author

The Potential Expertise Paradox in AI-Assisted Radiology.

Radiology·2026
Same author

Rhabdomyolysis and Acute Kidney Injury Associated With Acute Legionella Infection.

The American journal of case reports·2026
Same author

Radiomics-based Differentiation of Recurrent Brain Metastases from Treatment Effects: A Multi-Institutional Comparative Study with Advanced Imaging.

Radiology. Imaging cancer·2026
Same author

From Embeddings to Accuracy: Comparing Foundation Models for Radiographic Classification.

Journal of imaging informatics in medicine·2025
Same author

Comparative Evaluation of Radiomics and Deep Learning Models for Disease Detection in Chest Radiography.

Journal of imaging informatics in medicine·2025

Related Experiment Video

Updated: Jun 7, 2025

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

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

369

Technical Note: Neural Network Architectures for Self-Supervised Body Part Regression Models with Automated Localized

Michael Fei1, Alan B McMillan2

  • 1Creighton University School of Medicine, Phoenix, AZ, USA. Mkf55412@creighton.edu.

Journal of Imaging Informatics in Medicine
|November 13, 2024
PubMed
Summary
This summary is machine-generated.

Self-supervised learning models accurately identify body regions in medical scans, improving anatomical localization for segmentation tasks. EfficientNet demonstrated superior performance in predicting slice scores, enhancing downstream applications.

Keywords:
Artificial intelligenceDeep learningSegmentationSelf-supervised learning

More Related Videos

Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.0K
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

2.7K

Related Experiment Videos

Last Updated: Jun 7, 2025

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

Swin-PSAxialNet: An Efficient Multi-Organ Segmentation Technique

Published on: July 5, 2024

369
Deep Neural Networks for Image-Based Dietary Assessment
13:19

Deep Neural Networks for Image-Based Dietary Assessment

Published on: March 13, 2021

9.0K
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

2.7K

Area of Science:

  • Medical Imaging
  • Deep Learning
  • Computational Anatomy

Background:

  • Accurate body region identification is crucial for medical image analysis and deep learning pre-processing.
  • Traditional supervised learning methods require extensive manual labeling, which is time-consuming and costly.
  • Self-supervised learning offers a promising alternative by eliminating the need for labeled data.

Purpose of the Study:

  • To compare the performance of various neural network architectures in self-supervised body part regression (BPR) for anatomical localization.
  • To evaluate the effectiveness of BPR slice scores in developing anatomically localized segmentation models.
  • To assess the impact of BPR on the performance of downstream segmentation tasks.

Main Methods:

  • Implemented and compared VGG, ResNet, DenseNet, ConvNext, and EfficientNet architectures for BPR using the MONAI/Pytorch framework.
  • Calculated mean absolute error (MAE) by correlating landmark organ positions with predicted BPR slice scores.
  • Developed localized DynUNet segmentation models for thorax, upper abdomen, lower abdomen, and pelvis using BPR scores and compared Dice Similarity Coefficient (DSC) against baseline models.

Main Results:

  • EfficientNet architecture achieved the best BPR performance with a mean absolute error (MAE) of 3.18, significantly outperforming the VGG baseline (MAE 6.29).
  • Localized segmentation models demonstrated superior performance, achieving a Dice Similarity Coefficient (DSC) of 0.88 and outperforming baseline models in 16 out of 20 organs.
  • Enhanced neural networks, particularly EfficientNet, showed substantial improvements in localizing anatomical structures.

Conclusions:

  • Self-supervised learning with body part regression (BPR) is an effective strategy for anatomically localized segmentation in medical imaging.
  • The EfficientNet architecture offers a significant performance advantage for BPR tasks, leading to improved localization accuracy.
  • Utilizing BPR slice scores enhances the development of segmentation models, improving overall performance in downstream medical image analysis applications.