Search research articles

ABOUT JoVE

Overview Leadership Blog JoVE Help Center

AUTHORS

Publishing Process Editorial Board Scope & Policies Peer Review FAQ Submit

LIBRARIANS

Testimonials Subscriptions Access Resources Library Advisory Board FAQ

RESEARCH

JoVE Journal Methods Collections JoVE Encyclopedia of Experiments Archive

EDUCATION

JoVE Core JoVE Business JoVE Science Education JoVE Lab Manual Faculty Resource Center Faculty Site

Terms & Conditions of Use

Search research articles

Related Experiment Video

Updated: Nov 20, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

Training deep-learning segmentation models from severely limited data.

Yao Zhao^1,2, Dong Joo Rhee^1,2, Carlos Cardenas¹

¹Department of Radiation Physics, The University of Texas MD Anderson Cancer Center, Houston, TX, USA.

Medical Physics

|January 21, 2021

Summary

This summary is machine-generated.

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

Structure-property relationships in subnanometric transition metal tetramers.

RSC advances·2026

Same author

Artificial Intelligence in Image Assisted Radiation Oncology.

Cancers·2026

Same author

Quantifying the AI readiness gap: An international, multidisciplinary assessment of artificial intelligence literacy in the radiation oncology community.

Clinical and translational radiation oncology·2026

Same author

Use of Large Language Models to Enhance Failure Mode and Effects Analysis: A Case Study.

Advances in radiation oncology·2026

Same author

Feasibility of longitudinal relaxation rate mapping with non-Cartesian sampling and compressed sensing on a 1.5 T magnetic resonance linear accelerator.

Physics and imaging in radiation oncology·2026

Same author

Quantifying setup uncertainty between computed tomography guidance and magnetic resonance guidance in intramuscular metastases radiotherapy.

Journal of applied clinical medical physics·2026

This study introduces a novel data augmentation technique for deep learning segmentation models, enabling high-quality results even with limited contoured medical imaging data. The method effectively generates synthetic data to train segmentation networks for head and neck structures.

Area of Science:

Medical imaging analysis
Deep learning in radiology
Computational anatomy

Background:

Deep learning models require large annotated datasets for training.
Limited availability of contoured medical images poses a challenge for developing accurate segmentation models.
Accurate segmentation of head and neck structures is crucial for radiation therapy planning.

Purpose of the Study:

To develop a data augmentation strategy for generating high-quality deep learning segmentation models from a limited number of contoured cases.
To enable the creation of effective segmentation models with approximately 10 contoured cases.

Main Methods:

Deformable registration of 30 contoured CT scans to 200 unlabeled scans.
Training Principal Component Analysis (PCA) models on deformation vector fields to capture variations.

Keywords:

Auto-segmentation convolutional neural networks data augmentation deep learning principal component analysis

More Related Videos

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Published on: November 11, 2022

Related Experiment Videos

Last Updated: Nov 20, 2025

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography

Published on: November 30, 2022

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Deep Learning-Based Segmentation of Cryo-Electron Tomograms

Published on: November 11, 2022

Generating synthetic CT scans and contours using PCA models.

Training V-Net (3D convolutional neural network) using varying numbers of synthetic data and PCA models.

Evaluating segmentation performance using Dice similarity coefficients.

Main Results:

Achieved Dice similarity coefficients of 82.8% (right parotid), 82.0% (left parotid), and 74.2% (submandibular glands) using 2000 synthetic scans from 10 PCA models.
Results are comparable to state-of-the-art auto-contouring methods trained on significantly larger datasets.
Performance improvement plateaued beyond 10 PCA models or 2000 synthetic scans.

Conclusions:

Demonstrated an effective data augmentation approach for training high-quality deep learning segmentation models.
The method overcomes limitations of scarce annotated medical imaging data.
Enables the development of robust segmentation tools from limited contoured cases.