How to Minimize Data Loss in Multi-subject Studies with FSL

Symptoms:

When you have incomplete brain volume, as often happens, it is sometimes difficult to figure out which subjects account for the majority of the lost data. EvalMasks is a first draft tool to address this.

EvalMasks takes a series of mask files as input, and determines how much volume is removed from the total by each. It also outputs an image file for each input image showing (in gray) the area lost from the total volume by this image.

Note that the masks in question are found in the func_XXX.feat/reg_standard directories and are best renamed from their default name of mask.img.

What to do:

Example:

[Albert:]% EvalMasks Mask03.img Mask06.img Mask07.img Mask13.img Mask14.img Mask15.img

Mask03.img removes    657 of   97697 voxels, or  0.7%   --> Mask03vsAll.img
Mask06.img removes   4821 of  101861 voxels, or  4.7%   --> Mask06vsAll.img
Mask07.img removes   3550 of  100590 voxels, or  3.5%   --> Mask07vsAll.img
Mask13.img removes    204 of   97244 voxels, or  0.2%   --> Mask13vsAll.img
Mask14.img removes   6320 of  103360 voxels, or  6.1%   --> Mask14vsAll.img
Mask15.img removes   1215 of   98255 voxels, or  1.2%   --> Mask15vsAll.img

The program creates files called:: Mask03vsAll.img; Mask06vsAll.img; Mask07vsAll.img; Mask13vsAll.img; Mask14vsAll.img; Mask15vsAll.img

Below is the image data for Mask14vsAll.img from above:

Source code can be found at: EvalMasks.c

You can then take the statistical image data sets and merge them as an "Absolute Value Projection" using Mathalyze. This creates a single map with the hottest (and coolest) voxels from each input.

> Mathalyze -A FirstImage.img -B SecondImage.img -o MergedImage.img

Back to HowTo pages

This page is maintained by Mark Cohen [updated 3.18.03]