View in EDS HTML Full Text PDF Full Text

Split‐slice training and hyperparameter tuning of RAKI networks for simultaneous multi‐slice reconstruction.

Saved in:
Bibliographic Details
Title: Split‐slice training and hyperparameter tuning of RAKI networks for simultaneous multi‐slice reconstruction.
Authors: Nencka, Andrew S.1,2 (AUTHOR) anencka@mcw.edu, Arpinar, Volkan E.2 (AUTHOR), Bhave, Sampada3 (AUTHOR), Yang, Baolian4 (AUTHOR), Banerjee, Suchandrima4 (AUTHOR), McCrea, Michael5 (AUTHOR), Mickevicius, Nikolai J.6 (AUTHOR), Muftuler, L. Tugan2,5 (AUTHOR), Koch, Kevin M.1,2 (AUTHOR)
Source: Magnetic Resonance in Medicine. Jun2021, Vol. 85 Issue 6, p3272-3280. 9p.
Subjects: Network performance, Deep learning, Symptoms, Image reconstruction, Interpolation
Abstract: Purpose: Simultaneous multi‐slice acquisitions are essential for modern neuroimaging research, enabling high temporal resolution functional and high‐resolution q‐space sampling diffusion acquisitions. Recently, deep learning reconstruction techniques have been introduced for unaliasing these accelerated acquisitions, and robust artificial‐neural‐networks for k‐space interpolation (RAKI) have shown promising capabilities. This study systematically examines the impacts of hyperparameter selections for RAKI networks, and introduces a novel technique for training data generation which is analogous to the split‐slice formalism used in slice‐GRAPPA. Methods: RAKI networks were developed with variable hyperparameters and with and without split‐slice training data generation. Each network was trained and applied to five different datasets including acquisitions harmonized with Human Connectome Project lifespan protocol. Unaliasing performance was assessed through L1 errors computed between unaliased and calibration frequency‐space data. Results: Split‐slice training significantly improved network performance in nearly all hyperparameter configurations. Best unaliasing results were achieved with three layer RAKI networks using at least 64 convolutional filters with receptive fields of 7 voxels, 128 single‐voxel filters in the penultimate RAKI layer, batch normalization, and no training dropout with the split‐slice augmented training dataset. Networks trained without the split‐slice technique showed symptoms of network over‐fitting. Conclusions: Split‐slice training for simultaneous multi‐slice RAKI networks positively impacts network performance. Hyperparameter tuning of such reconstruction networks can lead to further improvements in unaliasing performance. [ABSTRACT FROM AUTHOR]
Copyright of Magnetic Resonance in Medicine is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)
Database: Engineering Source
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
Text:
  Availability: 1
Header DbId: egs
DbLabel: Engineering Source
An: 148886402
AccessLevel: 6
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Split‐slice training and hyperparameter tuning of RAKI networks for simultaneous multi‐slice reconstruction.
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Nencka%2C+Andrew+S%2E%22">Nencka, Andrew S.</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)<i> anencka@mcw.edu</i><br /><searchLink fieldCode="AR" term="%22Arpinar%2C+Volkan+E%2E%22">Arpinar, Volkan E.</searchLink><relatesTo>2</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Bhave%2C+Sampada%22">Bhave, Sampada</searchLink><relatesTo>3</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Yang%2C+Baolian%22">Yang, Baolian</searchLink><relatesTo>4</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Banerjee%2C+Suchandrima%22">Banerjee, Suchandrima</searchLink><relatesTo>4</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22McCrea%2C+Michael%22">McCrea, Michael</searchLink><relatesTo>5</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Mickevicius%2C+Nikolai+J%2E%22">Mickevicius, Nikolai J.</searchLink><relatesTo>6</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Muftuler%2C+L%2E+Tugan%22">Muftuler, L. Tugan</searchLink><relatesTo>2,5</relatesTo> (AUTHOR)<br /><searchLink fieldCode="AR" term="%22Koch%2C+Kevin+M%2E%22">Koch, Kevin M.</searchLink><relatesTo>1,2</relatesTo> (AUTHOR)
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="JN" term="%22Magnetic+Resonance+in+Medicine%22">Magnetic Resonance in Medicine</searchLink>. Jun2021, Vol. 85 Issue 6, p3272-3280. 9p.
– Name: Subject
  Label: Subjects
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Network+performance%22">Network performance</searchLink><br /><searchLink fieldCode="DE" term="%22Deep+learning%22">Deep learning</searchLink><br /><searchLink fieldCode="DE" term="%22Symptoms%22">Symptoms</searchLink><br /><searchLink fieldCode="DE" term="%22Image+reconstruction%22">Image reconstruction</searchLink><br /><searchLink fieldCode="DE" term="%22Interpolation%22">Interpolation</searchLink>
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: Purpose: Simultaneous multi‐slice acquisitions are essential for modern neuroimaging research, enabling high temporal resolution functional and high‐resolution q‐space sampling diffusion acquisitions. Recently, deep learning reconstruction techniques have been introduced for unaliasing these accelerated acquisitions, and robust artificial‐neural‐networks for k‐space interpolation (RAKI) have shown promising capabilities. This study systematically examines the impacts of hyperparameter selections for RAKI networks, and introduces a novel technique for training data generation which is analogous to the split‐slice formalism used in slice‐GRAPPA. Methods: RAKI networks were developed with variable hyperparameters and with and without split‐slice training data generation. Each network was trained and applied to five different datasets including acquisitions harmonized with Human Connectome Project lifespan protocol. Unaliasing performance was assessed through L1 errors computed between unaliased and calibration frequency‐space data. Results: Split‐slice training significantly improved network performance in nearly all hyperparameter configurations. Best unaliasing results were achieved with three layer RAKI networks using at least 64 convolutional filters with receptive fields of 7 voxels, 128 single‐voxel filters in the penultimate RAKI layer, batch normalization, and no training dropout with the split‐slice augmented training dataset. Networks trained without the split‐slice technique showed symptoms of network over‐fitting. Conclusions: Split‐slice training for simultaneous multi‐slice RAKI networks positively impacts network performance. Hyperparameter tuning of such reconstruction networks can lead to further improvements in unaliasing performance. [ABSTRACT FROM AUTHOR]
– Name: AbstractSuppliedCopyright
  Label:
  Group: Ab
  Data: <i>Copyright of Magnetic Resonance in Medicine is the property of Wiley-Blackwell and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract.</i> (Copyright applies to all Abstracts.)
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=egs&AN=148886402
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1002/mrm.28634
    Languages:
      – Code: eng
        Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 9
        StartPage: 3272
    Subjects:
      – SubjectFull: Network performance
        Type: general
      – SubjectFull: Deep learning
        Type: general
      – SubjectFull: Symptoms
        Type: general
      – SubjectFull: Image reconstruction
        Type: general
      – SubjectFull: Interpolation
        Type: general
    Titles:
      – TitleFull: Split‐slice training and hyperparameter tuning of RAKI networks for simultaneous multi‐slice reconstruction.
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Nencka, Andrew S.
      – PersonEntity:
          Name:
            NameFull: Arpinar, Volkan E.
      – PersonEntity:
          Name:
            NameFull: Bhave, Sampada
      – PersonEntity:
          Name:
            NameFull: Yang, Baolian
      – PersonEntity:
          Name:
            NameFull: Banerjee, Suchandrima
      – PersonEntity:
          Name:
            NameFull: McCrea, Michael
      – PersonEntity:
          Name:
            NameFull: Mickevicius, Nikolai J.
      – PersonEntity:
          Name:
            NameFull: Muftuler, L. Tugan
      – PersonEntity:
          Name:
            NameFull: Koch, Kevin M.
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 06
              Text: Jun2021
              Type: published
              Y: 2021
          Identifiers:
            – Type: issn-print
              Value: 07403194
          Numbering:
            – Type: volume
              Value: 85
            – Type: issue
              Value: 6
          Titles:
            – TitleFull: Magnetic Resonance in Medicine
              Type: main
ResultId 1