Contents
Step by Step Guide
First-Level FMRI
Run brain extraction on the structural image
make sure you name it with the same basename but ending with _brain (e.g. if the structural image is img005.nii.gz, then name the brain extracted version img005_brain.nii.gz)
- Processing of fieldmaps
- tight brain extraction of magnitude image (to avoid noisy voxels at the edge of the brain)
fsl_prepare_fieldmap (or alternative, if not Siemens data)
- find out values for echo spacing and phase encode direction
Open FEAT GUI (e.g. Feat_gui & on a Mac, or Feat & on other linux/unix)
Set the top left pull-down option to First-level analysis and the top right pull-down option to Full analysis (these are the defaults)
We will now work our way across the other tabs: Misc, Data, Pre-stats, Stats, Post-stats, Registration
Misc tab: this can be left as it is for now (we will revisit this later)
Data tab: load 4D data; change output directory name (optional); check that TR and number of volumes are correct; we will come back to set the filter later on
Pre-stats: the settings used here can vary from study to study but the following provides a basic default (see feat1 lecture for more info). Motion correction - on (mcflirt); B0 unwarping - recommended if a fieldmap was acquired (see separate instructions); Slice timing correction - off (temporal derivatives recommended instead); Spatial filtering - on (small amount recommended); Intensity normalisation - off (between-session normalisation is done automatically, but this within time-series form is not recommended); Melodic - off (turn on for additional exploratory analysis);
Stats tab: Select Use FILM prewhitening; click on Full model setup
- Within the General Linear Model window do the following:
On the EVs tab:
Set the Number of Original EVs appropriately
- For each individual EV tab (marked with the number 1, 2, …)
Set the EV name (something short but meaningful)
Set the Basic shape (the Custom 3 column format is the most commonly used)
If using the Custom 3 column format, prepare a plain text file (avoiding editors such as MS Word) containing one line for each stimulus associated with the condition to be modelled by this EV. Each line must contain three numbers: the onset time (in seconds); the duration (in seconds); the relative magnitude of each stimulus (usually set to 1)
To use the standard HRF the Convolution settings can be left as they are
Leave the Orthogonalise button unchecked (there are very limited circumstances when this can be validly used)
Leave the Add temporal derivative box checked (this is recommended for modelling slice timing and HRF timing delays)
Leave the Apply temporal filtering box checked
- For each individual EV tab (marked with the number 1, 2, …)
- Repeat the above for each EV
Select the Contrasts & F-tests tab:
Setup Contrasts & F-tests can normally be left at Original EVs (unless you use individual basis functions in your statistical question)
- Specify a name (title) and values for the individual contrasts
- At any point in the above process, press the View design button to see if what you've entered looks correct (and dismiss the image by clicking on it when you are done)
When you are finished, press Done (and dismiss the image after you've checked it)
- If there are any problems then you can re-edit the EVs and Contrasts by pressing the Full model setup button again
- Within the General Linear Model window do the following:
Post-stats tab: the defaults can be left as they are
Registration:
Check the Main structural image button and select the brain extracted structural image (defaults of Normal search and BBR are recommended)
- Nonlinear registration to standard space is recommended (select the Nonlinear button)
Misc tab: return to this tab and press the Estimate High Pass Filter then check that it has changed the High pass filter cutoff on the Data tab (note that this can be overly conservative sometimes, but should always provide a safe estimate)
When everything is set up correctly press GO (be patient - it should start a browser within a minute, but if not you can find the output feat directory and open the report.html file in a browser)
Note that you can use the Save button at any time to save your design, but it will be automatically saved for you in the feat directory when you press Go
- View the results:
- Look at the webpage output in the browser window
- View the result images that are in the feat directory (in FSLView)
- thresholded images are in the top-level feat directory
unthresholded images are in the stats subdirectory
registration results are in the reg subdirectory