Show simple item record

dc.contributor.authorHaskell, Melissa West
dc.date.accessioned2019-12-12T08:18:07Z
dc.date.created2019-05
dc.date.issued2019-05-13
dc.date.submitted2019
dc.identifier.citationHaskell, Melissa West. 2019. Retrospective Motion Correction for Magnetic Resonance Imaging. Doctoral dissertation, Harvard University, Graduate School of Arts & Sciences.
dc.identifier.urihttp://nrs.harvard.edu/urn-3:HUL.InstRepos:42029528*
dc.description.abstractMagnetic resonance imaging, or MRI, is an important tool for diagnosing disease and studying the human body. MRI provides exceptional soft tissues contrast compared to other medical imaging methods, and is essential for the diagnosis and treatment of many diseases. However, since its inception MRI has suffered from artifacts due to patient motion. Patient motion during MRI scans causes image artifacts that degrade diagnostic utility, often requiring repeated scans, patient callbacks, or lost diagnostic potential. Due to the prevalence and impact of motion in MRI, many techniques have been developed to detect and correct for patient motion, such as external tracking, MR motion tracking navigators, image entropy minimizations, and alternating optimizations. While all successful on some level, previous methods have sufficient weaknesses or side-effects to preclude their widespread clinical use. The goal of this work is to create a data consistency based retrospective motion correction method that can be translated to clinical use. Here we present two complementary methods for retrospective motion correction of brain MRI images: TArgeted Motion Estimation and Reduction (TAMER) and Network Accelerated Motion Estimation and Reduction (NAMER). TAMER solves for an uncorrupted MR image by jointly optimizing for the image and the unknown rigid-body motion of a subject’s head during the scan. We minimize the data consistency error of a SENSE+motion forward model, and use reduced modelling to make the optimization computationally feasible. NAMER builds on the TAMER method by incorporating a convolutional neural network (CNN) into the image reconstruction to accelerate algorithm convergence. TAMER simulation results show the improved optimization search direction accuracy using a reduced model and the potential for computational speedup. NAMER simulation results show how the inclusion of the CNN in the optimization allows for a truly parallel motion optimization to be constructed, which improves algorithm convergence and accuracy. TAMER phantom imaging results show the improvement in image quality over standard reconstructions when compared to a ground truth image. In vivo imaging experiments of healthy adult subjects and a neonatal subject show the potential of using TAMER and NAMER for motion mitigation in a clinical setting.
dc.description.sponsorshipBiophysics
dc.format.mimetypeapplication/pdf
dc.language.isoen
dash.licenseLAA
dc.subjectmagnetic resonance imaging
dc.subjectmotion correction
dc.subjectimage reconstruction
dc.subjectoptimization
dc.subjectmodel-based reconstruction
dc.subjectdeep learning
dc.subjectmachine learning
dc.titleRetrospective Motion Correction for Magnetic Resonance Imaging
dc.typeThesis or Dissertation
dash.depositing.authorHaskell, Melissa West
dc.date.available2019-12-12T08:18:07Z
thesis.degree.date2019
thesis.degree.grantorGraduate School of Arts & Sciences
thesis.degree.grantorGraduate School of Arts & Sciences
thesis.degree.levelDoctoral
thesis.degree.levelDoctoral
thesis.degree.nameDoctor of Philosophy
thesis.degree.nameDoctor of Philosophy
dc.contributor.committeeMemberAdalsteinsson, Elfar
dc.contributor.committeeMemberHuang, Susie Y.
dc.contributor.committeeMemberRosen, Bruce R.
dc.contributor.committeeMemberHogle, James M.
dc.type.materialtext
thesis.degree.departmentBiophysics
thesis.degree.departmentBiophysics
dash.identifier.vireo
dc.identifier.orcid0000-0002-7237-6731
dash.author.emailmwhaskell@gmail.com


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record