Methods for Function `dtiTensor' in Package `dti'

The method estimates, in each voxel, the diffusion tensor from the DWI data contained in an object of class "dtiData".

# S4 method for dtiData
dtiTensor(object, method=c( "nonlinear", "linear", "quasi-likelihood"),
          sigma = NULL, L = 1, mask=NULL, mc.cores = setCores( , reprt = FALSE))

Arguments

object	Object of class "dtiData"
method	Method for tensor estimation. May be "linear", or "nonlinear". method=="quasi-likelihood" solves the nonlinear regression problem with the expected value of the signal as regression function and weighting according to the signal variance.
sigma	(local) scale parameter of the signal's distribution.
L	(local) effective degrees of freedom.
mask	argument to specify a precomputed brain mask
mc.cores	Number of cores to use. Defaults to number of threads specified for openMP, see documentation of package awsMethods. Our experience suggests to use 4-6 cores if available.

Value

An object of class "dtiTensor".

Methods

obj = "ANY"

Returns a warning.

obj = "dtiData"

Estimate diffusion tensor from data in each voxel with the different options for the regression type and model for variance estimation. If method=="linear" estimates are obtained using a linearization of the tensor model. This was the estimate used in Tabelow et.al. (2008). method=="nonlinear" uses a nonlinear regression model with reparametrization that ensures the tensor to be positive semidefinite, see Koay et.al. (2006). The imlementation is based on R's internal C code for the BFGS optimization. method=="quasi-likelihood" solves the nonlinear regression problem with the expected value of the signal as regression function and weighting according to the signal variance. Tis requires additional parameters sigma and L characterizing the distribution of the signal. If varmethod=="replicates" the error variance is estimated from replicated gradient directions if possible, otherwise an estimate is obtained from the residual sum of squares. If varmodel=="global" a homogeneous variance is assumed and estimated as the median of the local variance estimates. sigma and 2*L are the scale parameter and degrees of freedom of the (local) signal distribution. L characterizes the effective number of coils. Both parameters are either scalars or arrays of the size of the images.

References

J. Polzehl and K. Tabelow, Beyond the diffusion tensor model: The package dti, Journal of Statistical Software, 44(12), 1-26 (2011).

K. Tabelow, H.U. Voss and J. Polzehl, Modeling the orientation distribution function by mixtures of angular central Gaussian distributions, Journal of Neuroscience Methods, 203(1), 200-211 (2012).

J. Polzehl and K. Tabelow, Structural adaptive smoothing in diffusion tensor imaging: The R package dti, Journal of Statistical Software, 31(9) 1-24 (2009).

K. Tabelow, J. Polzehl, V. Spokoiny and H.U. Voss. Diffusion Tensor Imaging: Structural adaptive smoothing, NeuroImage 39(4), 1763-1773 (2008).

C.G. Koay, J.D. Carew, A.L. Alexander, P.J. Basser and M.E. Meyerand. Investigation of Anomalous Estimates of Tensor-Derived Quantities in Diffusion Tensor Imaging, Magnetic Resonance in Medicine, 2006, 55, 930-936.

J. Polzehl, K. Tabelow (2019). Magnetic Resonance Brain Imaging: Modeling and Data Analysis Using R. Springer, Use R! series. Doi:10.1007/978-3-030-29184-6.

http://www.wias-berlin.de/projects/matheon_a3/

Author

Karsten Tabelow tabelow@wias-berlin.de
J\"org Polzehl polzehl@wias-berlin.de

See also

dtiData, readDWIdata, dtiIndices-methods, medinria, dtiData, dtiTensor dwiMixtensor

Examples

  if (FALSE) demo(dti_art)