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 Videos

MR-based synthetic CT generation using a deep convolutional neural network method.

Xiao Han1

  • 1Elekta Inc., Maryland Heights, MO, 63043, USA.

Medical Physics
|February 14, 2017
PubMed
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

Equal status in Ultimatum Games promotes rational sharing.

Scientific reports·2018
Same author

Preparation of Cyano-Substituted Tetraphenylethylene Derivatives and Their Applications in Solution-Processable OLEDs.

Molecules (Basel, Switzerland)·2018
Same author

Laser-Induced Conversion of Teflon into Fluorinated Nanodiamonds or Fluorinated Graphene.

ACS nano·2018
Same author

A Follow-up Study of Postoperative DCM Patients Using Diffusion MRI with DTI and NODDI.

Spine·2018
Same author

Technical Note: The impact of deformable image registration methods on dose warping.

Medical physics·2018
Same author

Red Blood Cells as Smart Delivery Systems.

Bioconjugate chemistry·2018

This study introduces a deep convolutional neural network (DCNN) for generating synthetic CT (sCT) images from MRI scans, improving radiotherapy accuracy and efficiency. The DCNN method significantly outperforms traditional atlas-based approaches in accuracy and speed.

Area of Science:

  • Medical Imaging
  • Radiotherapy
  • Artificial Intelligence

Background:

  • Magnetic resonance imaging (MRI) offers superior soft tissue contrast compared to computed tomography (CT) for radiotherapy.
  • Replacing CT with MRI in radiotherapy can reduce patient radiation dose and streamline clinical workflows.
  • Accurate CT-equivalent imaging (synthetic CT or sCT) is crucial for dose calculation and patient positioning in MR-only radiotherapy and PET-MR systems.

Purpose of the Study:

  • To develop and evaluate a novel deep convolutional neural network (DCNN) for generating synthetic CT (sCT) images from patient MR images.
  • To assess the performance of the DCNN method for brain tumor patients, comparing it against a conventional atlas-based approach.
  • To enable MR-only radiotherapy by providing accurate CT-equivalent data for dose calculations and patient positioning.
Keywords:
MRIconvolutional neural networkdeep learningradiation therapysynthetic CT

Related Experiment Videos

Main Methods:

  • A deep convolutional neural network (DCNN) with 27 convolutional layers and 35 million parameters was designed for end-to-end mapping from MR to CT images.
  • Transfer learning and pre-trained model weights were utilized to train the DCNN on a limited dataset of 18 brain tumor patients.
  • A sixfold cross-validation study was performed, comparing voxel-by-voxel accuracy of generated sCT against real CT images and an atlas-based method.

Main Results:

  • The DCNN method achieved a mean absolute error (MAE) below 85 HU for 13 out of 18 subjects, with an overall average MAE of 84.8 ± 17.3 HU.
  • The DCNN method demonstrated significantly better accuracy than the atlas-based method, with lower MAE (84.8 vs. 94.5 HU), mean squared error, and higher Pearson correlation coefficient.
  • Generation of a complete sCT volume using the trained DCNN model took only 9 seconds per patient, significantly faster than the atlas-based approach.

Conclusions:

  • A DCNN-based method for synthetic CT (sCT) generation from single-sequence MR images was successfully developed.
  • The DCNN method provides highly accurate sCT estimations in near real-time, outperforming atlas-based methods in both accuracy and speed.
  • Further validation on dose computation accuracy and larger patient cohorts is recommended, with potential for extensions to multi-sequence MR images.