From NITRC Wiki

Installing DOTS

System Requirements

DOTS requires an installation of the MIPAV software package (version 4.4 to 5.2 have been tested successfully), which is available on Windows, Mac, and Linux platforms. Additionally, DOTS can run within the JIST pipeline tool (version 2.0). The memory requirements for the DOTS plugin are dependent on the size of the data being processed. To safely run the plugin for a standard diffusion tensor brain data set with 1mm cubic resolution, we recommend that MIPAV is run with a 6GB virtual machine (in MIPAV, menu "Help > Memory allocation").

Installing MIPAV

MIPAV is available from http://mipav.cit.nih.gov/. Additional instructions for downloading and installing MIPAV with the JIST platform are available here.

Installing DOTS

Download the DOTS jar package and save it to a directory of your choice. Launch MIPAV and select "Plugins > Install plugin" from the MIPAV menu.

This will open the "Install Plugin" window. Click on "Browse" to select the directory where you saved the jar file, and select "Open". Then select the jar file, and click on the right arrow in the center of the window. Finally, click on "Install Plugins" at the bottom of the window. MIPAV will require some time to install the files. When installation is complete, a dialog will appear stating whether the plugins have been installed. Ignore any error message in this dialog box. Close any open MIPAV windows, then exit and relaunch MIPAV. When opening a new image in MIPAV, the "Plugins" menu should now list a sub-menu "Algorithms > DOTS".

Installing JIST (optional)

JIST can be downloaded from http://www.nitrc.org/projects/jist/. Once both DOTS and JIST are installed (and the JIST labrary has been re-loaded), JIST should automatically detect a DOTS module in "IACL > DTI". The DOTS module can then be inserted into any JIST processing pipeline.

Segmenting Diffusion Tensor Images with DOTS

DOTS infers a labeling of the major white matter tracts in a diffusion tensor image of the human brain based on a statistical atlas of shape and direction. The input data for the algorithm is the reconstructed tensor image, stored as a 4D image (dimensions X * Y * Z * T, with T=6). There are several possible conventions to store the diffusion tensor in a single array; DOTS currently supports "diagonal" (Dxx,Dyy,Dzz,Dxy,Dyz,Dzx), "upper triangular" (Dxx,Dxy,Dxz,Dyy,Dyz,Dzz) and "lower triangular" (Dxx,Dyx,Dyy,Dzx,Dzy,Dzz).

The DTI data is expected to be preprocessed to remove the extra-cranial structures, and the image should not include NaN values from the estimation of singular tensors (those should be set to zero). As an option, the algorithm can also input a mask of regions of disturbed white matter, for instance focal lesions, in order to improve the segmentation in these areas.

The results of the DOTS segmentation are:

• a pictorial 3D parcellation image, with a different color label for each tract, as specified in the atlas file. Regions of overlapping tracts are represented with a crosshatching pattern showing both labels alternatively. Note that because of the overlaps, this result is not to be used directly in any further processing.

• a 4D binary segmentation image, describing each tract separately. This image is the actual hard segmentation to use in further processing.

• a 3D soft membership image, showing the probability value for the segmented tract at each voxel. By construction, the probability for overlapping tracts is given as a whole and thus identical for both tracts. Probability values are not meaningful outside of the corresponding segmented tracts.