Old Quench Page

From VISTA LAB WIKI

Jump to: navigation, search

Contents

[edit] Tests

[edit] Voxel Offset Test

We came across some offset issue when converting from voxel space to ACPC space. A visual test and numerical test were developed to check if an offset is present

[edit] Visual Test

The visual test requires that a particular fiber passes through a point of interest. This can be done by creating a synthetic nifti image and creating a really small fiber in mrDiffusion. Below are two visual tests that were done In the first test I created a blank image with a white dot at the point of interest and checked if the fiber passes through that point. The blank image can be created by doing the following 2 operations

>DTIVolumeIO::loadVolumeNiftiNew()
 //Set image to black
 memset(data,0,sizeof(data[0])*((zdim-1)*xdim*ydim + (ydim-1)*xdim + (xdim-1)));
 // Point of intereset
 data[83 + 121*xdim + 61*xdim*ydim]=maxVal;
Visual voxel offset test #1

In the second image, the nifti image contains a straight line and the fibers were generated across that region.

Visual voxel offset test #2

[edit] Numerical Test

For a numerical test one can call the function MyFrame::DoesPointExist(DTIPathwayDatabase* pdb, float *trans, float *pt) which takes an input the pathway database, the transform vector from a DTIVolume and a point in voxel space. This function returns true if a matching point exists in the pdb otherwise it returns false. Note all the points in the pdb are translated by the trans vector first

[edit] User wishlist

  1. Adjust color for fiber bin 3 (yellow): currently it's too bright to discriminate different fiber trajectories. Maybe desaturate the hue or add some shading? Or maybe someone can tell me how to adjust the display properties of individual fiber group colors... --Davie 08:50, 11 March 2010 (PST)
Yellow2.png


Fiber group displayed as yellow -- for some reason, you can see more detail in the fibers in this screenshot than in the actual Quench environment.

Red2.png


SAME fiber group displayed as red, allowing user to get more information from the display

  1. View cortical surface mesh as a VOI
  2. Make 0 bin into a trash bin -- appending fibers thus get default assigned to a non-0 bin; when scoring, only the selected non-0 fibers are considered (and there is a way to only subtract from the current group)
  3. View any ROI in 3D in Quench
  4. Create a video showing how to interact with the more complicated functions, like "refining groups" and "VOIs"
  5. Be able to load multiple pdbs at once (without having to call the dialog several times)
  6. Be able to delete/remove/clear a pdb or selected fiber groups and load others (same background view, 0/trash bin empty)
  7. If pdbs are already added and you want to add a new pdb, append it to the fiber bin it was originally saved to and/or append it to an unassigned bin -- don't put it in 0/trash, because it mixes in with all the other trash fibers
  8. When fibers from different groups land on the same spot on a plane in the "surface" view, rather than deciding just to draw only the dots from one of the fiber groups, find some way to blend the colors to signify overlap -- this will make overlap easier to visually detect in the Quench environment. Bob suggests making the colors semi-transparent.

RFFALFFA CCoverlap.png
Screenshot showing how Quench currently decides to just draw only points from one fiber group in places on the plane where two different fiber groups intersect exactly

RFFA CC.png
Screenshot showing RFFA to CC endpoints alone, so that you can see where Quench decided not to show these endpoints

LFFA CC.png
Screenshot showing LFFA to CC endpoints alone, so that you can see where Quench decided not to show these endpoints

RFFALFFA CCzoomedout.png
Screenshot showing how the intersection drawing issue appears more dramatic when the slice is more "zoomed out"

[edit] Bug report

  1. When using the "refine selection" panel, even though the "trash" bin is not visible, the sliders start out including fibers I have already moved to the trash bin. If I first manually clean the fibers, and then want to use the "refine selection" panel to further trim the fibers, I don't want the fibers I have already trashed to be included again.

--Davie 17:01, 26 March 2010 (PDT)

  1. Errors when I try to load a .cst state file -- perhaps these errors would be moot when Quench can restore everything (loading the correct pdb files and backgrounds) when loading a saved state. But for now, here are the types of errors I'm running into.
  • In one outcome, I get the following error message: "ERROR: Number of state file pathways does not equal number currently loaded. No assignments restored!" -- even though I am loading the pdbs I originally edited, to the best of my knowledge
  • Loading proceeds, but yields crazy mixed up results (see screenshots) if fibers are not initially assigned to the exact bins they were originally assigned prior to any manual action in Quench. Here's what happens when I assign LFFA fibers to group 1 and RFFA fibers to group 2 -- given that the fiber group pairings were switched when first loaded in Quench prior to any manual action. It seems inconvenient to require me to memorize the order that I originally loaded the various pdb files in order for state file loading to proceed accurately.


Cst original.png
Screenshot showing fiber group status when "save state" was selected
Loadedcst differentorderofpdbs.png
Screenshot showing fiber group status when "load state" was selected --Davie 19:01, 13 February 2010 (PST)

[edit] Already fixed -- thanks Shireesh!

  1. After the 2/6/2010 updates, loading fibers appends to the first unassigned group, which is awesome -- but they don't seem to retain their filename, which I think was the case before the updates. Can this functionality be added/restored?
Personal tools