Diffusion Tools at the BIC

List of Topics/Programs

Acquisitions

We recommend the MGH DTI or DSI sequence, the parameters will depend on the application and the time you have available for acquiring data. The more directions you acquire, the more sensitive the acquisition is to subtle tract structure. You need to have sufficient SNR and/or spatial resolution to detect what you are looking for, and there is a tradeoff between these two. We have done between 2.0 and 2.8 mm isotropic resolution. If you are retrieving images that were transfered directly from the scanners at the BIC, the diffusion directions should be in the mincheader, in the 'acquisition:direction_{x,y,z}' fields. If you have dicom images, you can use dcm2mnc for the conversion (/usr/local/bic/bin/dcm2mnc on the BIC systems).

Transfer and Minc conversion

On scanner: export->BIC

a single minc file consisting of all b-values and directions sequentially in the minc time dimension (directions are fastest varying) will be accessible through the BIC systems in:

/data/transfer/minc/<patient name>

They will remain in that directory for one week after the transfer. If you need the dicom images, they will be in:

/data/transfer/dicom/<patient name>

for 48 hours after the transfer.

Getting the diffusion tools

Information on how to download and set up the tools can be found here.

Preprocessing

Adding diffusion encoding directions and b values to the minc header

This is currently done automatically in the minc conversion. However, if the files are from another site or are older acquisitions, the information might not be included. Either way, it is always good to check since this information will be used for all further processing. You can check what the directions are with: 
mincheader file.mnc | grep acquisition
For use with the dti.m matlab script and programs from mincdiffusion, the directions must be stored in the variables acquisition:direction_x, acquisition:direction_y, acquisition:direction_z, respectively, and the bvalues in acquisition:bvalues. The direction for a b value of 0 should be 0,0,0. The number of directions and b values must match the number of frames in the time dimension.

->To add diffusion directions and bvalues to the header, you can use: minc_modify_header-diff.pl. Please contact me if you would like a list of the diffusion directions for commonly used sequences.

Note: None of the minc tools (mincconcat, mincreshape, etc.) handle this header information correctly. Also, if you use mincmath, and you want to retain this information, use -copy_header.

Registration of DWI dataset with T1 anatomical

You should register each DWI dataset you have to the T1 anatomical, or at least to each other if registration with the T1 is not important for your study. Generating a brain mask is also useful for further processsing.
If you have a short set of DWIs, you can just transform the T1 to match the DWI set (using minctracc) and leave the diffusion encoding directions alone.

ALL in one

**The diff_preprocess script takes care of writing the diffusion information to the header (if it's not already there), registration of DWI sets to each other and to the anatomical, as well as brain masking. Please consult the manpage for more information on how we do this.

Tensor calculation

minctensor.pl is a wrapper for dti.m, which runs in matlab 6.0+, using emma or niak and minc utilities. 

minctensor.pl [options] <base_images_filename.mnc>
minctensor.pl -help for options or check the minctensor manpage

Multi-Fiber Reconstruction

Deterministic

mincDeconvolve is used to compute the orientation distribution function (ODF) using spherical harmonics decomposition. This allows for multi-fiber reconstruction within a voxel.

mincDeconvolve -help for options or check the manpage mincDeconvolve.

Probabilistic

mincProbDeconvolve like mincDeconvolve, is used to compute the orientation distribution function (ODF) using spherical harmonics decomposition. In addition, a cone of uncertainty is used to define the confidence of a particular direction of propagation.

mincProbDeconvolve -help for options or check the manpage mincProbDeconvolve.

Fibre tracking (FACT)

mincFibreTrack is mostly used for tracking with the standard FACT algorithm (which uses the prinicipal eigenvector direction for tracking) [1]. It also implements other tracking algorithms, such as the Monte-Carlo approach (using an arbitrary Orientation Distribution Function(ODF) for tracking),which are still WIPs at this point in time.
mincFibreTrack -help for options or check the manpage mincFibreTrack.

Visualization of fibre tracking results

This software will help you visualize tracts saved as vtk format (polydata) files. It also does a number of other useful things for displaying data in 3D.

minc3Dvis -help for options or check the manpage minc3Dvis.

Other useful tools

- Resampling a vector file (e.g. e* outputs of the minctensor algorithm) mincresample_vector.pl
- Concatenating diffusion volumes (along with diffusion info in header) concat-diff-header.pl
- Extract slab in time dimension of diffusion volume (along with diffusion info in header) extract-diff-frames.pl
NOTE: These programs are still WIPs, please feel free to help us work out any bugs/problems!

Other DTI processing packages

There are a number of other packages you can trying instead of using the above analysis tools by Jennifer Campbell. These include:

mincDTI (Andrew Janke - BIC)
FSL (Oxford)
DTI Studio (Johns Hopkins)
Camino (Manchester/London)
CINCH (Stanford)

Reference

The freely available, open source diffusion tools are provided without garantee and were developped by Dr. Jennifer Campbell. Please contact us if you use the tools for your study, in order to include the approriate reference. A standardized description of the methods you are probably using is available here methods:
Ilana Leppert
created: 2 Oct 2006
last modified:Dec 15th, 2011