This page contains the content of the educational course syllabus. For more information, go to the main ISMRM Educational content page.
Comparative review of available software packages
Speakers: Lindsay Walker and M. Okan Irfanoglu
In the world of diffusion imaging, there are a large number of choices when it comes to methods of data pre-processing and analysis. This is further complicated by the existence of many software packages each offering a different subset of tools for both pre-processing and analysis. It is not unreasonable to expect that the choice of pre-processing steps can affect the outcome of your analysis; although this is not well investigated and/or documented in the literature. In recent years this issue has started to reach mainstream. A few studies which have looked at the effects of different processing parameters on diffusion analysis are listed here:
1) a head-to-head comparison of different filter size use in voxel based analysis (Jones, DK, et al, The effect of filter size on VBM analyses of DT-MRI data, Neuroimage, 2005, Jun;26(2):546-54), which showed different regions of activation based on what filter size was selected for smoothing
2) the appropriateness of b-matrix reorientation post image registration (Leemans A. and Jones D.K. The B-matrix must be rotated when correcting for subject motion in DTI data, Magn Reson Med, 2009, in press)
3) analysis of the effect of a robust tensor estimation on DTI population analysis (Peterson, D, et al, The Impact of Robust Tensor Estimation on Voxel-Wise Analysis of DTI Data, proceedings of the 16th annual ISMRM, Toronto, 2008, p1823) and (Walker, L, et al, Regional Distribution of Outliers of Diffusion MRI in the Human Brain, , proceedings of the 16th annual ISMRM, Toronto, 2008, p139)
In this lecture, we will present a review of some commonly used software packages while highlighting some of the important pre-processing and analysis issues in the DTI field. When possible we will show a head-to-head comparison of the results of these different software packages on the same data in order to draw attention to the differences introduced by different techniques of pre-processing. A few of the issues that we will touch on include:
- Raw data import capabilities
Because of the lack of standardization in DTI data acquisition and storage, many software packages define their own data formats, their own tensor order and their own gradient coordinate framework.
- Gradient and b-value storage
Newly emerging DTI data acquisition protocols make use of different combinations of diffusion gradients and b-values, rendering the storage of this information more important. Support for multiple b-values and support for B-matrix in addition to gradient tables are the main focus points for this section.
- Diffusion weighted image (DWI) registration
The registration of DWI volumes is a crucial task for DTI processing. For this purpose we will determine which software packages have the capability for image registration, and describe any differences between registration techniques of the different packages.
- Image distortion correction
DWI data suffers from geometric distortions caused by eddy currents and Echo-Planar Imaging (EPI) susceptibility. Correction of these distortions is an important software feature that improves data quality.
- Dealing with artifacts in the data
Artifacts such as ghosting may render some portions of the data unusable. It is beneficial for a software package to provide a tool to reject these regions, such as volume rejection, slice rejection and/or voxel rejection.
- Tensor fitting algorithms
Most existing software packages provide linear least squares regression as the default tensor fitting approach. Non-standard tensor fitting algorithms such as non-linear fitting, constrained fitting, weighted fitting or robust fitting are newer tools that may be desired. Additionally, the presence of a goodness-of-fit measure, indicating how well the model explains the data, is an extremely useful tool.
- Scalar measures
Nearly every software package provides the functionality to compute basic scalar maps such as fractional anisotropy (FA) and apparent diffusion coefficient (ADC) for reporting diffusion tensor results. In some cases it may be desirable to have access to the tensor elements (Dxx, Dxy, Dxz, Dyy, Dyz, Dzz).
- ROI tools
ROI tools are an important part of medical analysis. The variety of these tools, ease of use, and the statistics provided about the ROI, are important software features.
- Tensor field registration and resampling
Tensor field registration and tensor resampling are new research topics that some software packages have started to incorporate as DTI tools.
- Fiber tractography
There are several important questions regarding fiber tracking;
- Does the software provide a fiber tracking method?
- What kind of tracking algorithm options are provided? (FACT, 2nd order Runge-Kutta, Tensorlines, etc.)
- Are any kind of fiber smoothing or fiber constraints provided?
- Is the fiber tracking performed online or offline; i.e. are all fibers computed at the beginning or are they computed on demand?
- What graphical aspects are provided for fiber visualization?
- Can statistics be computed based on fiber bundles or single fibers?
- Other visualization options
In some cases, voxel based visualization of tensors can be requested by the user. Glyph tools such as line-fields, tubes, ellipsoids or super-quadrics are features that software packages may provide to the user.
The following list of software packages will be reviewed as a sample to answer the issues listed above. We have provided links to the websites for these software packages for your reference.
- FSL (http://www.fmrib.ox.ac.uk/fsl/)
- Camino (http://www.cs.ucl.ac.uk/research/medic/camino/index.htm)
- DTIStudio (https://www.mristudio.org/)
- BrainVISA (http://www.brainvisa.info/)
- Slicer3 (http://www.slicer.org/)
- MedINRIA (http://www-sop.inria.fr/asclepios/software/MedINRIA/)
- TrackVis (http://www.trackvis.org/)
- Catnap (http://www.nbirn.net/tools/catnap/index.shtm)
- TORTOISE (http://www.tortoisedti.org)
- ExploreDTI (http://www.exploredti.com/)
This list of software packages is not intended as a comprehensive list of the numerous tools that are available to the DTI community. However, this lecture should provide insight into these most commonly used software packages. Additional information will be made available via the following website, including a more up-to-date list of software packages.