Mean magnetic susceptibility regularized susceptibility tensor imaging (MMSR-STI) for estimating orientations of white matter fibers in human brain.

Mark McIntosh,'s picture
PubMed URL:
van Zijl PC
Author List: 
Li X
van Zijl PC
Magn Reson Med
PubMed ID: 
An increasing number of studies show that magnetic susceptibility in white matter fibers is anisotropic and may be described by a tensor. However, the limited head rotation possible for in vivo human studies leads to an ill-conditioned inverse problem in susceptibility tensor imaging (STI). Here we suggest the combined use of limiting the susceptibility anisotropy to white matter and imposing morphology constraints on the mean magnetic susceptibility (MMS) for regularizing the STI inverse problem.The proposed MMS regularized STI (MMSR-STI) method was tested using computer simulations and in vivo human data collected at 3T. The fiber orientation estimated from both the STI and MMSR-STI methods was compared to that from diffusion tensor imaging (DTI).Computer simulations show that the MMSR-STI method provides a more accurate estimation of the susceptibility tensor than the conventional STI approach. Similarly, in vivo data show that use of the MMSR-STI method leads to a smaller difference between the fiber orientation estimated from STI and DTI for most selected white matter fibers.The proposed regularization strategy for STI can improve estimation of the susceptibility tensor in white matter.
Published Date: 
September, 2014

Bradley L. Schlaggar, M.D., Ph.D., Named President and CEO of Kennedy Krieger Institute

We’re thrilled to welcome Bradley L. Schlaggar, M.D., Ph.D., to the Kennedy Krieger family as our next President and CEO.

Learn more.

Appointments & Referrals



Read inspiring stories, news and updates about the Institute's patient care, research, special education, professional training, and community programs.


Resource Finder


A free resource that provides access to information and support for individuals and families living with developmental disabilities.