QDEC Feature List




GUI-less module to support usage with XNAT (dan marcus, tim olsen) and usage with Slicer3 (steve pieper and ron kikinis)


XNAT: support for local XNAT and client-server XNAT


be able to save analyzed data to a different directory, and load in analyzed data from a directory

volumetric analysis

build on Linux, Mac and Windows independent of freesurfer. (Windows requirement could be fulfilled by Slicer Qdec module).

get rid of the Subjects and Design panels. Instead, use a Wizard interface to load a table, check subjects, set the average subject name, choose measures, and specify other options. This will allow more room for GUI elements, as well as potentially have some context-sensitive help by dividing the steps into their own pages and having a "Help" button for each page.

add menu item to download XNAT result directly from website

update wiki page documentation, and create an 'advanced QDEC' page listing all the features not described in main tutorial

create wiki pages for XNAT-QDEC and Slicer-QDEC

dump pop-window error messages to the terminal, so that they can be copied and pasted

bind alt-q or ctrl-q to exit

Input / Subjects tab:


selection of SUBJECTS_DIR (default taken from enviro on startup)


selection of an average subject (default is fsaverage)


parse an input file of stats engine input parameters (qdec_glmfit)


put labels on the 'continuous factors' graph


allow the factor plot to be saved to postscript


connect 'exclude subject' option (right-click) to something that actually excludes the subject


allow the user to exclude subjects based on continuous factor range and discrete factor class


import aseg and aparc data from subjects stats files and create qdec.freesurfer.table.dat (whereby qdec automatically includes it contents in the continuous factors column)


error checking: check for existence of each subject in SUBJECTS_DIR


error checking: after a table is loaded, check that the expected number of levels for a factor was found, also check that there are no one-level discrete factors

show the 'nsubjects x nsubjects' matrix of dot products of thickness vectors

allow the user to zoom in on the graph contents

add Load Directory capability to load existing directory data

right click on a scatter plot point, and bring up a menu that allows that subject to be loaded with the appropriate ?h.white and aseg.mgz, then do a "send_spherical_point" to that subject with the clicked vertex and goto that point in tkmedit

allow multiple scatter plots, at least of the same data range (ie. left/right hippo volume)

Stats Engine / Design tab:


output files tarballable and able to load (saving all parameters used to create output)


fixup question generation to include ', accounting for <factor>'


display Degrees Of Freedom in the current design


allow selection of either DODS or DOSS in overall design

allow selection of either DODS or DOSS per factor (requires mri_glmfit change)

support for writing out the fsgd and contrasts without forcing an analysis

must work with more than two levels in a discrete factor, at least three

allow saving the design covariance matrix

put mri_glmfit shell output in the view frame

have a live search text field to find measures

display the design matrix in grid-block form

have a method to categorize measurement names. This could be done by using a naming scheme where categories are divided by a string token, e.g. Category|Name. The user would be able to hide or show all measures in a category in order to simplify the measurement lists.

allow loading a y.mgh (stack) file and run from it (instead of concatenating the data in the subjects dir). somehow need to make sure that the data in the stack is the same as the subjects in the qdec.table.dat file (and in the proper order). this feature is useful for fMRI, DTI and ASL files.

ability to correlate the cortical thickness of a particular vertex or the mean thickness of a region with the thickness of all other vertices in one hemisphere.

read file file called qdec.extensions which allows appending options to selection, namely allowing other surface measures, or other smoothing values

handle missing data (empty values)

de-mean a column of continuous data, to prevent the 'matrix is ill-conditioned or badly scaled' error

support for hemispheric interaction (need to add interhemispheric registration)

question text generated for DOSS analysis are wrong (not the same question as DODS)

rename y.mgh to ?h.y.mgh so that each hemi is distinguished

add -logy switch as option to mri_glmfit

(for rahul) parcellation glmfit: get a p-value for each parcellation unit, displayed either in bar-graph, or each unit as a single value

if the qcache files dont exist, then create them (run recon-all -s subj -qcache -target X -measure Y -fwhm Z)

have a smooth level between 10 and 15 (or have an entry box)

Contrast tab:

create a Contrast tab to break-apart Design matrix phase from Contrast matrix phase, allowing selection of particular contrast matrices (main effects, interactions, etc.)

display a list of selectable nuisance factors (the list of prior-selected discrete and continuous factors). In the Display panel, this would prevent any contrast question containing the text of that variable from showing up in the list

display the contrast and design matrices in grid-block form

GUI for defining the contrast matrix, raw entry with text showing regressors, use the form used in SPSS

Question generator - fill-in-the-blank

make mri_glmfit run in a separate thread

the design is retained, but qdec forgets what was a nuisance variable if you go back to the contrast after running an analysis

Display tab:


ROI selection (see Jenni's use case, #7)


optionally display alternative annotations


hide green-cross hairs (so that it is not visible when a snapshot is taken)


make sure Ctrl-left click always selects a point on the surface (as OpenGL point selector has no concept of our brain surface, and can pick outside points)


print fsgd plot as postscript


correcting for multiple comparisions: False Discovery Rate (FDR)


save vertex plot in postscript format


Why control-click instead of click?


Maybe the colorbar should be horizontal along the bottom?

correcting for multiple comparisions: clustering/GRF- at least generate a script which can be run

show residual FWHM

lock rotation to x,y,z axes (so that spinning the brain is easier)

allow saving the vertex plot data (as a csv file); same for ROI data

keep vertex plots a constant width, so snapshots are consistent (that is, don't change size from point to point to make room for varying label name widths)

in vertex plot, allow setting min and max axis values

in vertex plot, display R-value slope

mouse selection keys should match those in tksurfer (brucef)

display raw average thickness data, via 'mri_concat --i y.mgh --mean --o y-mean.mgh'

load and display gamma.mgh. Add a control to "View signficance of gamma" for a contrast, and mask the gamma to the contrast values over the current threshold

option to take a screen shot with the color bar

change the Display pane viewing area into a split pane. Move the plotting window into the bottom frame, so you can see it at the same time as the surface, or, alternatively, have two notebook tabs: surface and plot

enter number of degrees of rotation (default 90deg)

functionality to answer these questions:

-- is a particular parcellation significant compared to the rest of the brain?

-- where is the slope of a selected vertex significantly greater than other areas of the brain?

ability to select a region, and get a plot for that (that it, instead of just one vertex, plot the mean of the region)

the Regression Coefficients scalar is only the first frame: need to show all of them!

have a button to run the overlay on and off, so that you can see the underlying curvature

the snapshot tool doesn't work (it gives vTK errors)

The histogram is way too small. Maybe right-clicking on it gives you the option to bring up a big window?

show Talairach coordinates

Find max/min around the cursor (used to find max/min in a region)

Allow viewing R values

use mris_surfclsuter to set something like an 'Extent threshold' in order to display only clusters of a certain size and hide all clusters that are smaller

draw label using fill (of connected region), tksurfer does this

at each vertices, click to get ROC curves (like JMP), include sens, spec, neg/pos predictive value, neg/pos likelihood ratios (rahul)

option to view parcellations as wireframe, as opaque doesnt quite work

when plotting a scatter plot, dump the raw data to a text file so that someone can plot this in another package

drawing an ROI in the pial view maps it to a different spot in the inflated view; suggestion is for the ROI add and delete buttons to be grayed out when users are in the pial and white views so they don't attempt to draw anything there

QdecProject/FeatureList (last edited 2012-03-26 17:02:12 by NickSchmansky)