This stage creates the folder structure for your subject and timepoint. You can run this stage before you added the DICOM files in the corresponding RAWDATA folders. By modifying subject name and timepoint, you can quickly create the set of folders for your study.
Converts the raw DICOM data files in the RAWDATA subfolders into the Nifti format. It is possible to convert independently the structural data (T1 and T2), the diffusion (DSI, DTI, QBALL) and the functional data (rs-fMRI). Performs required flipping if needed. If the DICOM files have no file name ending, just enter * as the file pattern.
If you have not acquired any structural T2 image or rs-fMRI image, you can deselect the corresponding checkboxes. However, diffusion data and T1 data are required.
For the conversion, packages such as Diffusion Toolkit and Nibabel are used.
This stage allows to register the structural T1 image (where the parcellation of the cortical surface is extracted) onto the diffusion space.
Choose Linear or BBregister if you only acquired the structural T1 image and you miss the additional T2. In this way, the stage will try to align the T1 directly onto the b0 image.
If additional T2 images have been acquired, Nonlinear registration can be performed. A two steps procedure is performed:
We use the nonlinear registration approach in order to mitigate the nonlinear distortions which are present in diffusion images. Future versions of the pipeline will account for other methods to achieve this (e.g. fieldmaps).
The FSL linear registration tool FLIRT is used. Default parameters (which can be modified) generally give good results in most cases.
The FSL nonlinear registration tool FNIRT is used to perform this step. Default parameters (which can be modified) generally give good results in most cases.
The FREESURFER cross-modal registration tool BBREGISTER is used. Differently from FSL FLIRT, BBREGISTER registration algorithm exploits the FREESURFER segmentation results and can therefore be more robust in the context of this pipeline. Default parameters (which can be modified) generally give good results in most cases.
We use Freesurfer’s recon_all for the segmentation. You can provide custom parameters for recon_all.
We provide two parcellation schemes.
The Lausanne2008 parcellation is in experimental stage. Use it with caution. More information about the parcellation.
The registration transformations are applied to the white matter mask and the parcellation to map them into the diffusion space.
Use DiffusionToolkit for extracting the orientation distribution function (ODF), the default parameters are the same as DTKs.
You can also set parameters for the DTB_dtk2dir conversion. This can be helpful if you have to flip axes before tractography.
|--ix||invert x axis|
|--iy||invert y axis|
|--iz||invert z axis|
This module runs a classical streamline fiber-tracking algorithm (Weeden et al. (2003), Diffusion spectrum magnetic resonance imaging (DSI)) tractography adapted to deal with possible multiple directions inside each voxel.
The following parameters are automatically set by the mapper: Tracking is performed inside the white matter mask computed by FreeSurfer (–wm parameter) and is started in each non-zero voxel of the seed mask (–seed parameter); if no such a mask is give, then all voxels will be considered. Trajectories are iteratively propagated following coherent diffusion directions inside each voxel (–dir parameter) by using a fixed step size (–stepSize parameter) and the Euler integration method, and are stopped when a stopping criteria is met.
The following parameters are recommended to be explored: Stopping criteria are: trajectories are outside the white matter mask or no compatible direction are found below a specific crossing angle, –angle parameter). Only diffusion directions with a volume fraction greater than a threshold are considered (–vf parameter). For some imaging modalities, this parameter has no sense (e.g. DTI) and it is ignored. Multiple seed points can be created inside each voxel (–seeds parameter); this way, multiple trajectories will be started for every direction inside each voxel.
Apply cutoff filter: Fibers can be filtered depending on their length:
This stage merges the grey matter labeling and the tractography to create a connection matrix or brain graph for each resolution. A final tractography file is stored for each parcellation containing only fibers that start and end in grey matter regions.
Very general edge measures are used to construct the network, namely the number of fibers between two regions and their average length. Further measures can be computed using the Connectome Viewer using appropriate scalar volumes, tractography and label arrays.
This stage produces average time-courses for each cortical ROI, from resting-state fMRI (rsfMRI) data. FSL MCFLIRT is used to realign the rsfMRI time points and compute the mean rsfMRI volume. The T1 volume is then registered to the mean rsfMRI volume. It is possible to choose between two different linear registration tools: FLIRT or BBREGISTER (see ‘Registration’ step). The linear transformation T1-to-mean_rsfMRI is then applied to the cortical ROIs’ volumes corresponding to the selected parcellation scheme, and the averaged rsfMRI time-course is computed for each ROI.
The averaged time-courses are saved as Numpy matrices of dimensions number_of_ROIs X number_of_timepoints. Check the ‘Save .mat format?’ case if you wish to save the average time-courses in mat format too.
The output average time-series matrix can be suitably analysed through the Connectivity Decoding Toolkit (Richiardi J, Eryilmaz H, Schwartz S, Vuilleumier P, Van De Ville D. 2011. Decoding brain states from fMRI connectivity graphs. Neuroimage 56: 616-626). In this case, check the ‘Save .mat format?’ option, then import the matrices directly into Matlab Fand use them to feed the Brain Decoding Toolkit’ function ‘connectivityDecoding_filtering’.
Raw and processed data are stored in the connectome file for further analysis in the Connectome Viewer or elsewhere.
If you want to explore the pipeline parameters, beware that the output of the stages will be overwritten. Alternatively, you can duplicate the data folders.