This page provides a guide for a QIT workshop held as part of the USC Master's of Science in Neuroimaging program. The goal of the workshop is to provide an opportunity to work directly with neuroimaging data and to gain hands-on experience with various diffusion MRI techniques. The workshop doesn't require any prior programming experience, and it is designed to work on most laptops. That said, an interested reader can also use this as a starting point for more advanced scripting of image analysis tasks or integrating QIT in compute-heavy workflows with other datasets using the LONI Pipeline.
You can start by downloading the latest version of QIT here: http://cabeen.io/download/qit-latest.zip. If you haven't used QIT before, please follow the instructions listed on the Installation page. You may also benefit from checking out the Concepts pages linked on the left side panel, which will help you become familiar with the design and capabilities of QIT. Next, you should download the sample dataset: http://cabeen.io/download/qit-workshop-average.zip. You can then decompress the archive and keep it somewhere easily accessible.
The data included in this workshop was created by averaging a collection of MRIs from typical adults, which were acquired as part of the Human Connectome Project. The images and data were moderately downsampled from the full resolution to make the workshop run smoothly, but most anatomical features present in the original data can be found in this data as well. The archive contains the following groups of data:
models.dti: the diffusion tensor image volume (a Volume dataset)
models.xfib: the multi-fiber image volume (a Volume dataset)
models.noddi: the NODDI image volume (a Volume dataset)
regions: various atlas image labels (a set of Mask datasets)
tracts: various tractography reconstructions (a set of Curves datasets)
meshes: various surface models (a set of Mesh datasets)
masks: brain and white matter masks (Mask datasets)
Mask datasets are typically stored in the nifti format with the
Mask files typically have a
csv stored alongside them that indicates the name of each label. Diffusion models are stored as multiple volumes inside a directory, and while you can load the individual files, to do tractography or render glyphs you need to load the directory itself, e.g.
Curves data are both stored as VTK files with the
qitview should detect which is which based on the filename, but you should make sure the file loader correctly identifies the dataset type (the type can be changed using the combo box to the left of the filename). You can read more about the various dataset types here.
In this section, we provide instructions and video guides (no audio) for using QIT in various diffusion MRI tasks. You will only need the data provided in the above mentioned archive, and after the first step, the others are mostly self-contained. In total, the tutorial should take from 30-45 minutes to complete, but please feel free to take your time experimenting and to ask questions along the way!
Start Up and Loading Data
The first step will be to start
qitview and load some data. You should start by double clicking
qitview (on macOS) or
qitview.py (if you are on Windows). You should see a terminal window open that prints status messages followed by a graphical interface with several panels. You can load data by dragging data into the
qitview window (or by navigating through the File>Load Files menu item). You should load the file named
models.dti/dti_S0.nii.gz and try exploring the dataset. Tip: you can quickly change slices by hold Alt while clicking and dragging on an image slice.
Visualizing Surface Anatomy
Next, we'll examine some 3D models of anatomy represented by surfaces. You should load the data in the
meshes directory into the viewer, along with an image volume for reference. The surfaces show the boundaries of various anatomical structures. The larger structure is a white matter surface, and the smaller structures are subcortical nuclei. Tip: try to change the coloring of the subcortical meshes to the label attribute like in the video.
Visualizing Diffusion Glyphs
Next, we'll examine 3D rendering that depict diffusion modeling. You should load
models.xfib and follow the steps shown in the video below to create diffusion glyphs. Tip: watch for the step with a combo box incrementing from 0 to 1. This nudges the glyphs off the image slice so they are completely visible.
Visualizing Track Density
Next, we'll go over how to visualize fiber bundles in two ways. You should load a fiber bundle model, e.g.
tracts/arcuate.vtk.gz and a reference volume. You can see the bundle is represented by curves. The video shows how to convert it to a density map that can be viewed on top of the reference volume. Tip: try changing the colormap and opacity of the density map.
Comparing Diffusion Models
Next we'll create new tractography reconstructions and compare two different diffusion models. You should load two diffusion models,
models.xfib, along with a corpus callosum region mask
regions/fsa.ccwm/rois.nii.gz. The video shows how you can open the tractography module and create tracks for each of the diffusion models, and then observe the differences in the depicted anatomy. Tip: the toggle keyboard shortcut (Command-T on macOS or Control-T otherwise) will let you switch on and off the visibility of the highlighted datasets. You can also try changing the tractography parameters, e.g. increasing the samples or changing angle threshold.
Extracting Tracks Manually
Next, we'll show how you can extract tracks using a manually drawn seed mask. You should open a diffusion model volume and a reference dataset.
models.dti/dti_RGB.nii.gz is an efficient way to visualize white matter anatomy. You can also load atlas labels, e.g.
regions/jhu.regions/rois.nii.gz. You should create a new mask for seeding by right clicking on the model data and selecting the option shown in the video. You can then draw on the mask: first, select the mask in the Data Panel, then click the Edit tab in the Control Panel to see the options, then drag the mouse over the volume to draw the mask while holding down Alt+Control. Try drawing a region of interest on the corticospinal tract in the brainstem. Tip: you can erase a mask by dragging the mouse while holding down Alt+Control+Shift.
Filtering Tracks Interactively
Finally, we'll try interactively filtering tracks. You should load a reference volume and one of the fiber bundles, e.g.
tracts/arcuate.vtk.gz. You should create a sphere object as indicated in the video. You can then modify the sphere by first selecting it in the the Data Panel and then clicking and dragging the mouse while holding down either Alt (to move the sphere) or Alt+Shift (to resize the sphere). Then, you can filter the fiber bundle using the sphere by selecting the curves in the Data Panel and opening the Edit tab in the Control Panel. You should select the sphere as an include criteria as shown in the video. Now only the subset of curves that go through the sphere will be shown. You can similarly set the sphere to exclude curves, as shown in the video. Tip: the filtering only changes what is shown on the screen. If you want to save the filtered curves, you must either Export them or Retain them using the buttons in the Edit tab.
That's it, congratulations on finishing the workshop! Hopefully, you're now more familiar with a few common diffusion MRI tasks, and you have an idea of how to use
qitview for other tasks you may encounter in the future. Many of the tasks you completed in
qitview can be scripted or run on the command line using the
qit program. To learn more, please feel free to explore the rest of the documentation on this site, experiment the features in
qitview, or email me with any questions!