fMRI

In this example, we will process an fMRI data-set from a single subject acquired using a simple on/off block-design activation task.

1. Start the Dynamic Analysis tool, and select the General Linear Model tab.
2. Select the input image(s). In this example, each time point is in a separate image file, so we selected the button. Then we selected the button, and chose our input images. The file names are numbered in the right order, so the input images do not need to be reordered using the or buttons.
3. The time between images was set to 1.995 (which is the sequence repetition time), and we created an ROI file containing regions of interest that outlined the brain on every slice. This ROI file (which we called BrainMask.roi) was selected as the mask.
4. Since the subject moved slightly during the scan, we selected the check-box. We also provided a mild spatial blur to improve the SPMs by setting the smoothing filter FWHM to 3 mm.
5. The number of pre-steady-state images was set to 3, since the magnetisation has not reached a steady state for the first three time-points, making them brighter than the rest of the images.

6. The general linear model was set up to match the task. The task was a simple on-off activation paradigm, with 8 images of rest followed by 8 images or stimulation, and the whole cycle was repeated 8 times.

   • The option to specify the design in "scan number units" was selected. The single correlate was edited (press the button), choosing a "Repeating Block" design.

   • The correlate was given the name "MovingDots", since this was the stimulus. Both the rest duration and stimulus duration were set to 8 scans each, with the first period being a rest period.

   • On the "Confounds" tab, a correction for linear drift was selected, with the number of cycles for the Fourier basis functions set to 4. Since we have a total of 8 cycles during the full stimulus/rest period, this will not filter out the required activation pattern.

   If you want to use this setup repeatedly, or use it in batch processing, you can save it to an XML disk file by pressing the button 7. We set a p-value of 0.05 for the significance level, and select the Bonferroni correction check-box so that we can be more confident that we are seeing real activation in the T-statistic SPMs. 8. The output image base name was set to "Test".

The picture below shows the full set-up.

Pressing the button results in the creation of several output images, with name created from the base name "Test" set above, and the correlate name "MovingDots". These are:

• TestMovingDots a map of the amplitude of the β value of this correlate.
• TestMovingDots-p a map of the p-value for the test that the amplitude of this correlate is significantly greater than zero.
• TestMovingDots-T a map of the T-value for the test that the amplitude of this correlate is significantly greater than zero.
• TestRMSE a map of the root-mean-square error between the fitted model and the data.
• Testresiduals the residuals over all time points (apart from the pre-steady-state time points) after fitting of the GLM. This was created because we had the check-box selected. You would not normally want to see this for fMRI analysis.

Below, you can see the T-statistic map overlaid on one of the input images for a visual stimulation task.

T-statistic activation map overlaid on an fMRI input image.