Estimating 3D Deformable Motion from a series of Fast 2D MRI Images Public Deposited

Downloadable Content

Download PDF
Last Modified
  • March 21, 2019
Creator
  • Brown, Jason
    • Affiliation: School of Medicine, UNC/NCSU Joint Department of Biomedical Engineering
Abstract
  • In this application, we estimated patient-specific 3D deformable motion in the abdomen from a series of fast 2D images. CLARET (Correction via Limited-Angle Residues in External Beam Therapy) is an image registration method that has been used to estimate 3D deformable motion from 2D X-ray images. This work generalizes CLARET and extends it to use with MRI images of the abdomen. Using CLARET to predict the 3D motion of a subject from a set of 2D projection images has the potential to be used in fast MRI imaging of dynamic processes. The method begins with acquisition of a 4D respiratory-gated image set using a gradient-echo sequence. From the 4D set, a patient-specific motion model was derived, as well as a regression relationship between the 3D anatomy and 2D slice images taken with a specific geometry. The second dataset was a series of fast 2D gradient-echo images of the same subject, which are used via the regression relationship to estimate the 3D body poses at each time point. Before testing on the acquired 2D dataset, CLARET was tested on a simulated dataset which confirmed the method accurately predicted random warps of the dataset. In a free breathing experiment, the CLARET procedure gave motion estimates that reduced alignment error mean and variance in the 2D frames. We conclude that CLARET can be applied in an MRI setting and produces fast instantaneous motion estimates with less registration error than a time-averaged estimate.
Date of publication
Keyword
DOI
Resource type
Rights statement
  • In Copyright
Advisor
  • Gomez, Shawn
  • Lalush, David
  • An, Hongyu
Degree
  • Master of Science
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2017
Language
Parents:

This work has no parents.

Items