dtiprocess:Dtiestim

There are several methods to estimate the tensors which you can specify with the option --method lls-wls-nls-ml . Here is a short description of the different methods:

lls Linear least squares. Standard estimation technique that recovers the tensor parameters by multiplying the log of the normalized signal intensities by the pseudo-inverse of the gradient matrix. Default option.

wls Weighted least squares. This method is similar to the linear least squares method except that the gradient matrix is weighted by the original lls estimate. (See Salvador, R., Pena, A., Menon, D. K., Carpenter, T. A., Pickard, J. D., and Bullmore, E. T. Formal characterization and extension of the linearized diffusion tensor model. Human Brain Mapping 24, 2 (Feb. 2005), 144-155. for more information on this method). This method is recommended for most applications. The weight for each iteration can be specified with the --weight_iterations. It is not currently the default due to occasional matrix singularities.

nls Non-linear least squares. This method does not take the log of the signal and requires an optimization based on levenberg-marquadt to optimize the parameters of the signal. The lls estimate is used as an initialization. For this method the step size can be specified with the --step option.

ml Maximum likelihood estimation. This method is experimental and is not currently recommended. For this ml method the sigma can be specified with the option --sigma and the step size can be specified with the --step option.

You can set a threshold (--threshold) to have the tensor estimated to only a subset of voxels. All the baseline voxel value higher than the threshold define the voxels where the tensors are computed. If not specified the threshold is calculated using an OTSU threshold on the baseline image.

dtiestim also can extract a few scalar images out of the DWI set of images: the average baseline image (--B0) which is the average of all the B0s. the IDWI (--idwi)which is the geometric mean of the diffusion images. You can also load a mask if you want to compute the tensors only where the voxels are non-zero (--brain_mask) or a negative mask and the tensors will be estimated where the negative mask has zero values (--bad_region_mask)

• Brain Mask
  • Brain Mask [----brain_mask] [-M]: Brain mask. Image where for every voxel == 0 the tensors are not estimated.
  • Bad Region Mask [----bad_region_mask] [-B]: Bad region mask. Image where for every voxel > 0 the tensors are not estimated.
  • Baseline Image [----B0] : Baseline image, average of all baseline images.
  • IDWI [----idwi] : idwi output image. Image with isotropic diffusion-weighted information = geometric mean of diffusion images.
  • DWI Image [----dwi_image] : DWI image volume (required)
  • Tensor Output [----tensor_output] : Tensor OutputImage.
  • Method [----method] [-m]: Esitmation method (lls,wls,nls,ml) Default value: lls
  • Step Size [----step] [-s]: Gradient descent step size (for nls and ml methods) Default value: .00000001
  • Verbose [----verbose] [-v]: produce verbose output Default value: 0