Peristimulus Plots

Introduction

SPM and FSL are powerful tools for analyzing fMRI data. However, the statistical maps most people generate with these tools can be difficult to interpret. Generating peristimulus plots can allow you to get a better idea of what your data actually looks like, and can help you determine if a region shows an increased amplitude of activity or a more sustained response to a stimuli. To generate peristimulus plots you will need:
  1. MRIcron developmental release.
  2. A 4D fMRI dataset (typically motion corrected and smoothed)
  3. A FSL format 3-column text file for each condition you wish to analyze.
  4. Optional: regions of interest for specific brain regions.
You can also click here to download a sample data set (14mb).

Basic Usage

Here are step-by-step instructions
  1. launching MRIcron. Then choose '4D traces' from the View menu.
  2. Press the 'Open Data' button.
  3. MRIcron will now display a timeline for your data. If you have loaded multiple regions, a separate line displays each ROI. If you have not selected any ROIs, you will be shown the currently selected voxel - use MRIcron's main window to select a voxel you want to view and then press the red refresh button in the timeline window to see the timeiline for this voxel. Note that if you have loaded any event onsets, each condition is shown as a unique color of vertical stripes - for example in the example left hand taps are shown as red bars and right taps are shown as green bars. Note with the example datasets that left taps are followed by increases in signal for the right ROI, while right taps are followed by increasing signal in the left hemisphere.
    Timeline
  4. Before generating peristimulus plots, make sure that the TR is accurately set. Our sample data has a TR of 3 seconds, and this is correctly reported in the image file, so MRIcron correctly reports a TR of 3 seconds. If your TR is incorrect, the events will not be correctly aligned with your images.
  5. Press the 'Plot' button to generate phase-locked peristimulus plot. You will want to check the settings for your peristimulus plot
    1. The bin width sets the resolution for plot - smaller bins are more precise but noiser. By default, the bin width is set to your TR, in our example 3 seconds.
    2. The pre-stimulus bins sets the number of baseline bins. In our example we are setting 4 bins (12 seconds).
    3. The number of post-stimulus bins plot signal changes after an event has been presented. Remember that fMRI signals are sluggish, and take 5-6 seconds to peak. For the example, set this to 14 (42 seconds).
    4. If you slice time corrected your data, check the appropriate box. Event times will be adjusted for the acquisition of the middle-slice in your volume (e.g. all of your onsets will be adjusted by 0.5 TR).
    5. The save peristimulus volume button allows you to save a separate 3D dataset for each time bin. This is an advanced feature we will discuss later.
      Settings
  6. MRIcron generates a peristimulus plot. Different colors are used for the different conditions, while different line styles are used for the different regions of interest. For our example, note that the right hemisphere shows a response for left but not right taps, while the reverse is true for the left hemisphere. The peak amplitude is about 1% signal change. While this effect sounds small, note that we are averaging over a large number of voxels which in this case were selected baseed solely on anatomy, rather than post-hoc selecting the single most active voxel. Also note that the error bars are rather small.
periplot

Advanced Usage


Notes

My software gives you a direct view into how your data looks. Also note that a single timepoint can be averaged into a number of bins (e.g. if the events occur rapidly, one scan could show a timepoint which is after a previous event but before one or more others). Furthermore, my software does not attempt to remove data from other conditions. An alternative approach is to fit each condition and then plot the data having regressed out the variability explained by other conditions. A nice implementation of this alternative approach is described here.

logo