• DocumentCode
    2638068
  • Title

    A closed-form method for improving inter-subject coherence in diffusion tensor magnetic resonance imaging

  • Author

    Lange, Nicholas ; Jones, Derek K. ; Pierpaoli, Carlo

  • Author_Institution
    Dept. of Psychiatry & Biostat., Harvard Univ. Sch. of Medicine & Public Health, Boston, MA, USA
  • fYear
    2004
  • fDate
    15-18 April 2004
  • Firstpage
    1506
  • Abstract
    A simple method is presented to reduce within-group inter-subject scatter in diffusion tensor magnetic resonance imaging (DT-MRI). By "borrowing strength" across co-registered subjects to accommodate indirect effects of unmeasured machine and physiological noise, the method reduces voxel-specific tensor variance across subjects. The technique may aid in fiber bundle atlas construction, in testing differences between groups of subjects, and in automated outlier detection. While the technique does not in itself address DT-MRI signal artifact issues directly, it may serve to lessen the effects of these artifacts when their sources have not been measured. An example application to DT-MRI of twelve healthy male volunteers at the splenium of the corpus callosum slightly right of midline demonstrates the possible utility of the method.
  • Keywords
    biomedical MRI; brain; image registration; medical image processing; tensors; DT-MRI signal artifact issues; automated outlier detection; closed-form method; corpus callosum; diffusion tensor magnetic resonance imaging; fiber bundle atlas construction; inter-subject coherence; physiological noise; splenium; unmeasured machine; voxel-specific tensor variance; 1f noise; Anisotropic magnetoresistance; Brain; Diffusion tensor imaging; Eigenvalues and eigenfunctions; Humans; Image analysis; Least squares approximation; Magnetic noise; Tensile stress;
  • fLanguage
    English
  • Publisher
    ieee
  • Conference_Titel
    Biomedical Imaging: Nano to Macro, 2004. IEEE International Symposium on
  • Print_ISBN
    0-7803-8388-5
  • Type

    conf

  • DOI
    10.1109/ISBI.2004.1398836
  • Filename
    1398836