Index

# Name

func2roi-sess - this is a tool for ROI analyses

# Synopsis

func2roi-sess -roidef <roidefname> -analysis <analysisname> -label <labelname> -s <subjectname>

# Arguments

none

## Required Flagged Arguments

 -roidef name name of ROI definition -analysis name source is averaged data from analysis -sesslabel name use name.label found in sessid/labels or -anatlabel name use name.label found in SUBJECTS_DIR/subject/labels or -labelfile file give full path of label file -s sessid or -sf sessidfile subject name or name of text file containing list of subjects -d srchdir or -df srchdirfile name of subject directory

# Description

func2roi-sess is an FS-FAST program for averaging functional time courses across select spatial voxels, the Region of Interest, or ROI. Currently, there is only support for averaging the results created by selxavg-sess. The key to understanding func2roi-sess is to understand the many ways to select the voxels which constitute the ROI. Broadly speaking, the ROI can be constrained both structurally and functionally. For example, the user may want to constrain the ROI to be within fusiform gyrus. However, there may be many voxels in this structure that are not activated by the task at hand, and averaging the non-active voxels with the task-related voxels will water down the significance of the ROI. Thus, func2roi-sess allows the user to specify a minimum level of functional activation before a voxel can be included in the ROI. Note that care must be taken not to evaluate the significance of the ROI using the same functional criteria which which the ROI was chosen as such a statistic will have elevated false-positive rates. In what follows, the structural constraint will be referred to as the label. The functional constraint will be referred to as the functional mask. The intersection of these two regions will be referred to as the final mask.

## Specifying the Structural Constraint

The structural constraint is specified using a label file. This file can be created by tkmedit or tksurfer and always has the extension .label. The label file is just a list of the xyz coordinates of 1 mm^3 voxels to include in the label. These coordinates can be defined in either a subject's native anatomical space or in talaiarch space, the latter being more convenient when the same label will be used across multiple subjects. func2roi-sess is designed to easily accommodate either method.

A per-subject label file can be specified in one of two ways, depending upon where the file is stored. It can be stored within the functional session directory in a subdirectory called labels; this type is specified with -sesslabel command-line flag followed by the label name. The label file can also be stored in the subject's FreeSurfer anatomical directory in a subdirectory called label; this type is specified with -anatlabel command-line flag followed by the label name. Note that the label name SHOULD NOT include a .label extension.

If a single label is to be applied across all sessions, use the -labelfile command-line flag followed by the full path of the label file (WITH the .label extension).

If the label was constructed in talairach space, then include the -labelspace tal option. Note also that a label does not have to be purely structural. For example, one can display a functional map when choosing label voxels with tkmedit or tksurfer, and only tag those voxels that are active. If using this method, take care not to evaluate the significance of the ROI with the same criteria used to select the voxels. Note also that a structural criteria need not be specified at all.

Each label point occupies 1 mm^3 of space whereas each functional voxel typically occupies about 45 mm^3. This creates a question as to how many label points must fall into a voxel for that voxel to be considered part of the label. By default, a voxel will be included in the label if any label point falls into it. This can be helpful when creating labels by hand because only a few points need to be specified (ie, it allows for sparse labels). On the other hand, if two labels are adjacent to each other, the the voxels on the border will end up belonging to both labels. To avoid this, users can specify the minimum fraction of a voxels volume that must be filled with label points for that that voxel to be considered part of the label (this is equivalent to setting a minium number of label points that must fall into a voxel). This fraction is specified with the -labelfillthresh flag followed by the fraction to use (eg, -labelfillthresh 0.5). The fraction should be between 0 and 1 (0 is equivalent to not having a label constraint). Setting the threshold to 0.5 assures that a voxel cannot belong to more than one label. Note, however, that this can have undesirable effects when the label is defined on the surface because there might not be enough surface points in a voxel to reach threshold.

## Specifying the Functional Constraint

The basis of the functional constraint is usually a contrast map as computed by stxgrinder-sess. The contrast can be obtained from an individual session or the average of a group of sessions. If a group is used, the contrast from a fixed or random effects averaging can be used, and the results from one analysis/contrast can be used to create an ROI for another analysis. The per-voxel threshold is chosen using the -maskthresh flag followed by -log10 of the significance threshold (eg, a threshold of .01 would require -maskthresh 2). A voxel can also be included or excluded based on the sign of the significance at that voxel using the -masktail flag followed by either abs for absolute value (ie, ignore sign), pos to include only voxels with positively signed significances, or neg to include only voxels with negatively signed significances. stxgrinder-sess typically produces three significance maps: sig, minsig, and fsig, any of which can be used for the functional mask using the -maskmap option followed by the name of the map. For event-related designs, the sig map will have multiple frames, and a particular frame can be chosen using the -maskframe option followed by the zero-based frame number. By default, func2roi-sess will search for the contrast within the analysis for which the ROI is being computed (ie, the argument of the -analysis flag). The contrast from a different analysis can be used by specifying the -maskanalysis flag. Finally, if the contrast is based on a group average, the group name (as supplied to isxavg-fe-sess or isxavg-re-sess) is specified using -masksessid. Note that it is not necessary to have a functional mask at all.

# Outputs

When func2roi-sess is run, it creates a new directory in each session under the analysis subdirectory for that session. The name of this directory is the argument passed with the -roidef flag. There will be several files within this directory representing the results of the ROI analysis. There will be a file called h\_000.bfloat. This is where the hemodynamic averages (as computed by selxavg-sess) averaged within the ROI are stored. There will also be h-offset\_000.bfloat in which the average functional value within the ROI is stored. Note that these are the same types of files found in the analysis directory and in the resampled (ie, tal and sph) directories. In fact, one can think of ROI analysis as a type of resampling, and many of the FS-FAST programs that can be applied to resampled spaces (eg, stxgrinder-sess, sliceview-sess, isxavg-fe-sess, and isxavg-re-sess) can be applied to the ROI. The ROI volume is actually just a single voxel (ie, one slice with one row and one column). In addition to the ROI averages, there is a volume called mask which has the same dimension as the original functional volume. A voxel in this volume has a value of 1.0 if that voxel was included in the {\em final mask} and a value of 0.0 otherwise. Thus the final mask can be viewed with any FS-FAST program used to view a contrast just by specifying "-contrast" roidef (roidef is the name given to the ROI in ) and "-map" mask. The minimum threshold should be set to 0.5 when viewing. Programs that can be used to view the mask are: sliceview-sess, tkmedit-sess, and paint-sess/[" surf-sess"]. Viewing the final mask can be helpful in assuring that the structural and functional constraints result in an ROI that is close to expectations.

## Getting Results in ASCII Format

Often, one wants to pool all the ROI averages across sessions into a single table so that they can be read into a statistics package for more elaborate analysis. The results from func2roi-sess can be collected into such a table using roisummary-sess. See roisummary-sess.

# Examples

## Example 1: Per-Session Label and Functional Mask

Let's say that you have two sessions, s1 and s2, for which you have already run individual event-related analyses (main2) as well as a contrast (allvfix). The analysis has 2 conditions, with TER=TR=2sec tPreStim=4sec, and Time Window=18sec. The allvfix contrast gives you an indication of which voxels are activated by the overall task. A neuroanatomist has graciously labeled the fusiform and striate regions of each subjects' brain giving you a total of 4 label files. You create a directory in s1 called labels and put the fusiform.label and striate.label for subject 1 into this directory. You create a directory in s2 called labels and put the fusiform.label and striate.label for subject 2 into this directory.

### Subsection a: Fusiform with Positive Activation

func2roi-sess -roidef fusi-avf-pos-2 -analysis main2 -sesslabel fusiform \
-s s1 -s s2 -d . 

This will create an ROI named fusi-avf-pos-2 which uses the fusiform.label in each subjects' labels directory along with the allvfix contrast as the functional constraint. Only voxels inside the pre-defined fusiform label that exceed a significance of 10^-2 in the minsig map and are positively correlated with the task will be selected. Once the final ROI is determined from the functional and structural constraints, func2roi-sess will average the ROI voxels from the main2 analysis together, and create a directory called fusi-avf-pos-2 under main2 (in each session).

At this point, there are four things you can do next: (1) pool the data into a text file, (2) view the hemodynamic response in the ROI for each session, (3) view the final mask, or (4) inter-subject average.

Pooling the data into a text file

Here you will run roisummary-sess to create a text file of all the ROI data for both session 1 and 2. See roisummary-sess

View the hemodynamic response

To view the hemodynamic response for this ROI for session s1,

sliceview-sess -s s1 -d . -analysis main2 -space fusi-avf-pos-2 -slice 0

This will bring up what looks like one big voxel. Click on it and hit 'g' to view the hemodynamic response.

The final mask can be visualized with either sliceview, tkmedit, or tksurfer.

To view in sliceview, type:

sliceview-sess -s s1 -d . -analysis main2 -contrast fusi-avf-pos-2 \
-map mask -slice mos -thmin .5 -thmax 1

This will bring up a mosaic in which the base image will be the average functional value and the voxels selected for the ROI as a yellow overlay.

To view in tkmedit, type:

tkmedit-sess -s s1 -d . -analysis main2 -contrast fusi-avf-pos-2 \
-map mask -fthresh .5 -fmid 1

This will bring up the ROI overlaid onto the 3D anatomical for that subject.

To view in tksurfer, type

paint-sess -s s1 -d . -analysis main2 -contrast fusi-avf-pos-2 -map mask
surf-sess -s s1 -d . -analysis main2 -contrast fusi-avf-pos-2 -map mask \
-fthresh .5 -fmid 1

This will paint and display the final mask on the surface.

### Subsubsection b: Fusiform with Negative Activation

The first part of example 1 showed how to obtain voxels in the fusiform label that were positively correlated with the task. Here, we show how to use the same label and contrast to define a new ROI in which the activity in the voxels is negatively correlated with the task:

func2roi-sess -roidef fusi-avf-neg-2 -analysis main2 -sesslabel fusiform \
-maskcontrast allvfix -maskthresh 2 -masktail neg -maskmap minsig -s s1 -s s2 -d . 

Note the option -masktail neg instructs func2roi-sess to use the negatively correlated voxels, and we have changed the name of the ROI to fusi-avf-neg-2. This will create a directory called fusi-avf-neg-2 under main2 in each session. The remainder of Part 1 applies to Part 2; all you need to do is substitute fusi-avf-neg-2 for fusi-avf-pos-2.

## Example 2: Cross-Session Label

Your friendly neuroanatomist has left you for some other brain-mapper because you insist on using sluggish, command-line-driven functional analysis tools. But you need one more ROI. You'll have to draw it yourself. You know about where Broca's area should be in the Talairach brain, so you pull up the Talairach brain in tkmedit:

tkmedit talairach T1

You can create the label by pressing the middle mouse button. The marked voxels will appear green. When finished, pull down the File menu and select Save Label As. Hit the little folder icon, and save the label in your study directory as broca.label. Exit out of tkmedit. Cd to your study directory and make sure there is a file called broca.label.

func2roi-sess -roidef broca-avf-pos-2 -analysis main2 \
-labelfile broca.label -labelspace tal -maskcontrast allvfix \
-maskthresh 2 -masktail neg -maskmap minsig -s s1 -s s2 -d . 

Note that -sesslabel fusiform has been replaced with -labelfile broca.label (also note the .label extension) and the addition of -labelspace tal.

## Do I have to have a structural mask and a functional mask ?

No. You have to have one or the other, but both are not required. However, it probably won't make much sense to just have a functional mask as any and all areas in the brain that meet the functional critria will be included.

None