【TrackVis 教程】基于DTI的纤维追踪及显示

NOTE: For regular users who would always use the GUI front-end, you do not have to understand the details of the following workflow. But having an idea of the process is recommended.

To understand how Diffusion Toolkit's GUI works, it would be necessary to know the actually workflow behind the scene. The workflow of Diffusion Toolkit (DTK) contains 4 major steps (some of them are optional).

1. Convert the raw DICOM images into a series of ANALYZE images. The command-line program to do that is diff_unpack. If the raw images are not DICOM, you will have to convert them to ANALYZE manually following the rules below:

• Each analyze image must contain only one timepoint (or volume, as we often refer to. or one gradient direction, in the case of diffusion imaging).
• Filenames of the images must be named with a prefix followed by number, such as xxxxx001.img, xxxxx002.img ...
• Low B (B0) images must be on top. For example, if you have 60 diffusion-weighted images with 5 b0 images, you image filenames should be like this: xxxxx001 to xxxxx005 are b0 images, xxxxx006 to xxxxx65 are high b images.

2. Reconstruct the raw images.

• For DTI, dti_recon does the job, given the corresponding gradient table. The results are a set of tensor images, DWI, ADC and FA maps.
• For DSI, odf_recon is used directly to generate a odf file and a max file, as well as a few maps. At this point our DSI recon program only takes pre-calculated recon matrices, corresponding to pre-defined gradient tables.
• For HARDI/Q-Ball, users are allowed to provide a customized gradient table (just like for DTI). hardi_mat must be called first to generate the corresponding recon matrix, then odf_recon takes that recon matrix and reconstructs raw images to odf and maps.

3. Fiber tracking. Takes the reconstructed data and generated a track file.

• For DTI, dti_tracker is used.
• For DSI/HARDI/Q-Ball, odf_tracker is used.