FEAT 3 Practical

Practical Overview

This is the third FEAT practical. It leads you through some of the more advanced usage and concepts in both single-session and higher-level FEAT analysis.

Outlier Inference

In the last practical we have run the group analyses with the outlier de-weighting option turned off. In this section we will look at the output that is generated when outlier de-weighting is performed. Note that outlier de-weighting can be performed for any of the mixed effects analyses, but is not applicable to a fixed effects analysis. In the interest of time, we have provided you with the stripped down results of a group Feat analysis that has been setup and run with outlier de-weighting for you.

The experiment is looking to infer the group mean effect over a group of 18 subjects. The effect of interest is the BOLD response in the feedback phase of a decision making experiment. Move into the group Feat directory corresponding to this lower-level contrast:

cd ~/fsl_course_data/fmri/decmak/decmak.gfeat/cope1.feat

Inside the stats directory are the results of the higher level run with outlier inference turned on. Take a look at the files in that directory:

ls stats

The prob_outlier1.nii.gz file is a 4D niftii file giving the probability that each subject has outlier data on a voxelwise basis. The global_prob_outlier1.nii.gz file is a 3D niftii file that indicates the proportion of the subjects classified as outliers at each voxel (see the lecture for more details). Note there are versions of these files for each variance group in the analysis. In this case, there was only one variance group. Load the global_prob_outlier1 file into Fslview.

fslview stats/global_prob_outlier1.nii.gz

Use File->Add Standard to load in the MNI standard brain MNI152_T1_2mm_brain. Use File->Add to load in the prob_outlier1 file, and also load in the filtered_func_data in the ~/fsl_course_data/fmri/decmak/decmak.gfeat directory. Note that the filtered_func_data file is a 4D niftii file of the 18 subjects' first-level effect sizes (COPEs) at each voxel, in other words the "data" that gets used to calculate the group average at each voxel.

Select the prob_outlier image and select Tools->Timeseries. Then de-select the prob_outlier image so that it does not get displayed on the spatial map. Do the same for filtered_func_data.

Navigate around the brain to see where outlier behaviour occurs, and see how this relates to the data in the filtered_func_data timeseries plots and to the probability that each subject has outlier data in the prob_outlier1.nii.gz timeseries plots. Which subject is causing the most outlier behaviour?

Inside the stats directory is the file zstats1_no_outliers. These are the z-stats resulting from the higher level analysis run with the outlier inference turned off (note that this would not normally be available when you have run outlier inference - it is provided for you here to allow you to make comparisons). Use File->Add to load the z-stats into Fslview for the analyses with (~/fsl_course_data/fmri/decmak/decmak.gfeat/stats/zstat1) and without (~/fsl_course_data/fmri/decmak/decmak.gfeat/stats/zstat1_no_outliers). Take a look at how the zstats are changed when there are outliers detected.

Contrasts in Parametric Designs

We can use contrasts in FEAT to investigate different levels of stimulation. In the following simulated-data example we have 3 different stimulation heights: low, medium and high, and we want to know if there is a linear or quadratic trend relating stimulation strength to response strength in different areas of the brain.

cd ~/fsl_course_data/fmri/art2 Feat &
Select the input data artdatamultistims.nii.gz (note that because the data has only 3 slices a pop-up appears to notify you that the default settings in FEAT have been changed - this does not concern us, as we are going to turn off pre-stats and registration in this analysis).
This dataset has already been preprocessed, so change Full analysis to Stats + Post-stats.
Now setup the experimental design. The experimental design is as follows.
There are 3 EVs (one for each stimulation level) and all are boxcar designs with ON time of 20 seconds and OFF of 60 seconds, but with different phases:
EV1: low stimulation, phase 60 seconds
EV2: medium stimulation, phase 40 seconds
EV3: high stimulation, phase 20 seconds
Note that it is not important for this example whether you use temporal derivatives or not.
Next, setup three contrasts to look for the three different stimulations separately.
answer
Now add more contrasts to also allow us to investigate positive and negative linear trends.
answer
Setup the contrasts to look for positive and negative quadratic trends.
answer
What would the contrast be to look for a linear trend when there are six stimulation strengths?
answer
Give all the contrasts suitable titles and press Done.
Turn off all registration by selecting the Registration tab and making sure that the buttons on the left are all deselected.
Now press Go; wait for FEAT to finish and view the results. Identify which areas of the brain have the different types of trend AND in which directions.

Temporal Derivatives

We add temporal derivatives into the model to account for small delays (in either direction) between the model and the data. We can view this delay; it may be physiologically interesting. By applying the following commands you can manipulate the different parameter estimate values to get a delay image. We will use the audio-visual dataset results that we have already obtained in the first FEAT practical session:

cd ~/fsl_course_data/fmri/av/fmri.feat/stats

Now use fslmaths to calculate the delay. Make sure that you mask the delay by the activation - it doesn't make sense to show delay where the signal looks nothing like the model.

hint

answer

View the output: this is in units of % of TRs. It seems that there is a fairly consistent shift between the visual model and the data. Reopen the webpage report in a web browser, for example

open ~/fsl_course_data/fmri/av/fmri.feat/report_poststats.html

You can see this lag in the first timeseries plot as a temporal displacement between the data plot and the partial model fit; that partial model fit is the fit due just to EV1 (as COPE1 was [1 0]). Now look at the full model fit curve; because the temporal derivative (EV2) is included in the full model fit, it has done a good job of shifting the complete fit to better match the data.

Interactions

In this section we will look for an interaction effect in the audio-visual dataset:

cd ~/fsl_course_data/fmri/av

Type Feat &
Press Load and select the design.fsf file in the fmri.feat directory. This loads up the design we used in the earlier practical - saving us the effort of having to setup many of the same options again.
We have already preprocessed the data so we do not want to unnecessarily repeat pre-stats. To make this so, change to Stats + Post-stats, AND change the 4D data (in the Data section) to be filtered_func_data.nii.gz from inside the fmri.feat directory (produced from the previous run). The filtered_func_data.nii.gz file is the 4D data that was previously saved in the FEAT output directory when pre-stats was performed.
Go to Stats and Full model setup and increase the number of EVs to 3. Select the third EV and change the Basic shape to Interaction. The defaults are fine.
Now go to the contrasts and setup two extra t contrasts to look for positive and negative interactions: [0 0 1] and [0 0 -1]. Give the contrasts some helpful Titles. Press View Design and Efficiency. Study the design matrix and covariance/eigenvalue plots - you can see from the Effect required values that the interactions should be found as efficiently as the primary effects.
To save time turn off all registrations.
Now press Go, wait for it to finish and view the results. Note where the FEAT output directory has been created - with the name derived from the input data. Identify if there are any areas of the brain which have positive or negative interactions.

HRF Basis Functions

This section shows you how basis functions can be setup and used in FEAT. The dataset we will use is a jittered single-event experiment with 200 time points. The stimulus is heat applied for 3 seconds with an average inter-stimulus interval of 70 secs. We will only analyze one slice to allow for quick processing.

cd ~/fsl_course_data/fmri/bf

Type Feat &
Press Select 4D data and select filtdata.nii.gz. This dataset has already been preprocessed, so change Full analysis to Stats + Post-stats.
Set the High pass filter cutoff to 50sec.
Setup one EV using the Custom (3 column format) file pain_paradigm.txt. Note that this is the underlying experimental stimulus and is assumed to also correspond to the neuronal response to the stimulus. This needs to be convolved with our assumed HRF.

To start with we will analyze the dataset assuming a fixed Gamma HRF (no basis functions) and then compare the results with a set of the optimal linear basis functions.

Hence, for the EV leave the Convolution as Gamma, BUT turn OFF Add temporal derivative as we want to show the results for when a simple, fixed HRF is assumed.
Take a look at the contrasts - the default OC1 contrast of [1] is fine.
Press Done and view the design.
Take a look at the Post-stats tab - the defaults are fine.
Turn off all registrations. We are only processing one slice and are happy to observe the results on a low resolution EPI.
Press Go and wait for the results. On the FEAT report page look at the time series plots of data vs model.

We will now process the same data using FMRIB's Linear Optimal Basis Set (FLOBS) and compare the results.

Type Feat &

Press Load and select the design.fsf file in the filtdata.feat created from the fixed Gamma HRF analysis we have just performed. This loads up the design we just used - saving us the effort of having to setup many of the same options again.

The only bit we need to change is in the Stats tab. We need to change the Convolution option to Optimal/custom basis functions.

We can now specify the FMRIB's Linear Optimal Basis Set (FLOBS) filename we wish to use. By default the FLOBS provided in $FSLDIR/etc/default_flobs.flobs is used - so leave that selected. You can generate your own customised FLOBS by selecting Utils->Make_flobs on the main FEAT setup Gui (if you have time later then come back and try generating your own basis functions with this GUI and feed that into FEAT).

Now look at the Contrasts and F-tests tab. We have two ways of setting up contrasts with basis functions. This is via consideration of the "Original EVs" or the "Real EVs".

The default is to work with the "Original EVs". The "Original EVs" correspond to the true underlying effects/conditions in the experiment. Hence, to investigate if the effect is significant, a simple OC1 contrast [1] suffices.

Now press View Design and the actual design matrix to be used is displayed. This shows one EV for each of the basis functions convolved with the experimental stimulus. The single "Original EV" is automatically expanded into the "Real EVs" that need to be actually used in the design matrix when modelling with basis functions. Note that the OC1 [1] contrast will be actually carried out by the F1 f contrast. Why is this the correct way to do the test?

Return to Contrasts and F-tests and change Original EVs to Real EVs. The single "Original EV" is again automatically expanded into the "Real EVs" actually used in the design matrix, but now you have control of the actual contrast settings used.

Again press View Design and the design should be exactly the same as it was before.

Imagine that you want a contrast to test for a negative effect on an Original EV. What would this look like as a contrast on the Real EVs? Set one up and switch between Original and Real EV mode in the GUI to see if you are right. Once you've done this revert to your original contrasts before proceeding.

Press Done and then press Go and wait for the results. Take a look at the FEAT report. In particular, look at the peristimulus plots by clicking on the model fit plots. The peristimulus plots show the fits to a single epoch of stimulation. The data actually acquired at each repeat of the stimulation is plotted as a scatter of points. The "full model fit" on the peristimulus plot indicates the HRF formed by fitting the basis set to the data. Compare this FEAT report with the FEAT report from the single HRF Gamma run previously. Are there any differences?

If you get time at the end of this session try running a range of different basis sets, e.g. Sinusoidal, FIR. In particular, have a go at making your own customised basis set (FLOBS) using Utils->Make_flobs on the main FEAT setup GUI.

Contrast Masking

One problem with using Basis functions is that we have to use f contrasts to look for a significant effect. These are inherently two-tailed, hence we can not tell the difference between "activation" and "deactivation".

An unsophisticated way to remove "deactivation", for example, is to use Contrast Masking to mask the f test with the t contrast for the largest basis function EV (C1). Rerun the analysis, but this time performing this contrast masking (this is setup from the Post-stats tab on the Feat GUI; you will want to use the Mask using (Z>0) option). How do the results with and without this contrast masking differ, and how do the results compare with the single gamma HRF now?

If you click on the F-stat time course you get the FEAT Time Series Report for the f-statistic. One useful piece of information available here is how many voxels passed thresholding for the f-test. It can be found on the Full model fit - averaged over all active voxels, weighted by Z-stat values plot. Although the images look similar, you can see that there is a reduced number. If you have time, find the two relevant cluster mask images and use fslmaths to subtract them to easily find the differences.

This is the end of FEAT session 3.