Getting high-quality eye tracking is hard, especially when combined with other techniques like MEG or fMRI.
Here a few tips for improving eye-tracking quality in neuroimaging environments that you won't find in a manual,
though make sure to read your manual first (e.g., this one ).
By default, I am referring to the EyeLink eye tracker, but many of the tips apply to other systems as well.
I plan to update this list from time to time. If you have additional points or comments, please reach out.
Tip (1): Adjust the illuminator angle. When struggling to get good calibration, many people play around with the camera first. While this can help indeed, more often than not, the problem is not the camera, but the illuminator (i.e. the infrared light source). Especially on long-range setups, changing the illuminator angle relative to the participant can help a lot.
Tip (2): Adjust the illuminator power. When the illuminator is too strong, diffuse reflections can occur that confuse the eye tracker (e.g., reflections on the MRI headcoil). When it is too weak, everything drowns in noise. Try out a few different illuminator-power settings before your next study.
Tip (3): Do not rely on auto-threshold. Adjust the pupil and corneal thresholds manually after auto-thresholding. Especially increasing the corneal threshold beyond auto-threshold often helps a lot.
Tip (4): The eyes should move during thresholding. While setting the pupil and corneal thresholds, ask participants to move the eyes across the screen. Do not set the thresholds at center fixation or you risk losing the pupil as soon as the eyes move during the experiment.
Tip (5): Match the screen luminance. When setting the pupil and corneal threshold, and during calibration, make sure the luminance of the screen approximately matches the one of your experiment. As pupil size depends on luminance, your thresholds may not be ideal for your stimuli otherwise.
Tip (6): Keep the eye slighty out-of-focus. Many people think the eye needs to be perfectly in focus of the camera. This is not always true. In the presence of high frequency noise, typical for many MRI environments, keeping the eye slightly out of focus can help stabilize the tracking by smoothing the image. Don't over-do this obviously, but give it a try. I learned this trick during an official EyeLink training.
Tip (7): Use duct tape to redirect reflections. Unwanted reflections (e.g., on the MRI headcoil or the MEG) can confuse the eye tracker. The fix is simple! Use some Duct tape to change the reflections. Done!
Tip (8): Never touch the mirror. On many setups, the eyes are tracked via a mirror display. Be aware that those are not regular mirrors! On a regular mirror, the reflective surface is protected by glass. This is not the case for the mirrors typically used for the eyelink and other eye trackers. Each time you touch them, you damage the reflective foil. Be careful!
Tip (9): Remove fake eyelashes. If your participant has fake eyelashes, ask them to remove them. They often mess up the tracking big time.
Tip (10): Lower the frame rate. Maybe voodoo, but lowering the framerate to 500Hz or lower seems to make the tracking more robust in some cases. Can anyone else confirm this?
Tip (11): Use drift correction. In both MEG and MRI, head position can drift over time (e.g., when the neck muscles relax). Consider performing drift correction more often (e.g., every 10 minutes or so, it takes about 5 seconds).
Tip (12): Offer more breaks. One of the main killers of good eye tracking data is fatigue. Design your experiment such that the participant can rest their eyes every few minutes. This is especially important for fixation paradigms.
Tip (13): Monitor the eye tracker. Once the experiment is running, keep checking the eye tracking computer for problems. Sometimes, you can adjust things a tiny bit even though the experiment has already started. Worst case, re-start the run.
Tip (14): Avoid prospective motion correction. This advice may not apply to all systems, but a few prospective motion correction systems use infrared light with a similar wavelength as the EyeLink. This can cause massive artifacts, so better turn it off.
Tip (15): Try MR-based eye tracking. If you cannot get decent camera-based eye tracking inside the MRI, or if the data has been collected already, consider trying out DeepMReye .