Day 1 · Acquisition, preprocessing, and the GLM
Introduces Statistical Parametric Mapping, its history and why it is widely used, then begins the preprocessing pipeline with the two steps people most often skip: looking at the raw images and reorienting them to resemble the SPM EPI template, and running ArtRepair to find and replace bad volumes. These steps give later registration the best possible starting point and produce covariates for the first-level model.
Take-aways
Key terms
Outline
Hands-on
The walk-through below is distilled from the course cheat sheets. Data paths refer to the course Dropbox folder (e.g. data/auditory_oddball).
Mind fMRI Image Acquisition and Analyses Course - Cheat Sheet for Course Re-orientation
-Re-orientation – using the ‘Display Routine’ in SPM12.
Purpose:
Have a look at your data, check to make sure image is in the right orientation; voxel sizes etc are all correct. Look for artifacts. This is likely the only time you will look at your images.
Adjust your data so that it looks like the SPM EPI template. This will provide SPM with the best starting estimates for your data. The more similar your data are to SPM templates, the more likely the processes will run smoothly.
Initiate:
– Display function.
Start spm12.
Click on the ‘Display’ button.
‘select the first image in the time series for subject 1, run 1.
Select directory ‘c:\mind\data\auditory_oddball\AOD_raw\s01\run1’
use the filter settings [1:200] in the window to help you select the right scans
Click and select ‘s01_aod_run1.nii,1’
CLICK on ‘DONE’ to pull up the first image.
It is often useful to review the EPI template image on Display as well. This picture is available from the lecture slides on spatial normalization.
You may manually alter the way that spm views the image. For s01/run1 (and s01/run2 below), please enter the following:
In the boxes to the left, you can ‘reorient’ the images so they look like the template. The below values in x, y, z, will move the images into closer register. Then after the images look better, you can click ‘Reorient’ button botton left.
Then select all images for this subject (you may choose to do this for all runs for this subject) and click done. Then display the images again and click ‘origin’ and you will see that SPM now ‘knows’ details about your data and SPM will use this information to help spatial registration steps (like normalization).
Select data for subject 1 session 1 - directory ‘c:\mind\data\auditory_oddball\AOD_raw\s01\run1’
use the filter settings [1:200] in the window to help you select the right scans
Right click on mouse and ‘select all’ images
s01_aod_run1.nii,1 to s01_aod_run1.nii,189’
- CLICK on ‘DONE’
Repeat for different runs/subjects.
You may be prompted to save the matrix file for the changes you made, please feel free to save this file as a record of your edits. Or writing them down in your codebook.
This process will create ‘s01_aod_run1.mat’ (a matlab file that contains the transformation/re-orientation info), it will also modify the ‘s01_aod_run1.nii’ file/header fields. And if you saved the reorient values, then there will also be a ‘s01_aod_run1_reorient.mat’.
Mind fMRI Image Acquisition and Analyses Course - Cheat Sheet for Art Repair (a process to find and destroy bad images in your time series!).
Under toolboxes, select ‘Art Repair’ in SPM.
- 
This GUI will emerge, we will be using the ‘Bad volumes, detect and repair’. Click that button.
Purpose:
To identify bad images due to motion or other MRI artifacts (i.e., gradient issues, RF spikes).
To replace bad images with interpolated images and then output a covariate (zeros for all time points except a 1 where the bad images is). These covariates are used at the first level to address the bad image (that has been replaced).
Initiate:
After you click on Bad volumes: detect and repair, a dialog box will populate the bottom left window in the SPM gui. ‘Which global mean to use? - select Auto. Then it asks whether you have ‘realignment files’ – click no. Next it asks to input the images to analyze. Go select ‘…AOD_raw/s01/run1/ and then expand s01_aod_run1.nii. Highlight all images and select. Click done. Now you have the final question, ‘always repair 1st scan of each session’ Note that you have already reviewed the first scan of the session during ‘reorientation’ so there should not be any problems with this image, so normally you can select ‘no’. but it’s fine to also select ‘yes’.
The coded will now review your data.
This screen will pop up.
This will highlight whether there are any bad images. In this case there are no bad images. If you did see flagged images, then click ‘Repair’ and the code will write out a new volume of images for you where the bad images have been replaced and new vectors will be outputted as well that contain the covariate you will include in the first level model.
After clicking ‘repair’ select ‘interp’ which will provide interpolated images, replacing the bad images. This will generate a new nii file ‘vs01_aod_run1.nii’. the ‘v’ (volume repair) represents the new fixed image file. Also, you will see ‘art_repaired.txt’ (contains image number that were repaired), ‘art_regressors.txt’ (vectors containing 0s for all images in time series and a 1 where the repaired image was in the time series; one regressor per repaired image; these vectors should be imported into spm first level analyses for this subject; ‘artglobaldataart_repair_example.jpg – copy of the output from art repair.
From the instructors' research
Examples of these concepts in published work by the course instructors.
Day 1 · Acquisition, preprocessing, and the GLM
