Eye movement during a magnetic resonance imaging (MRI) scan correlates with brain activity. As such, researchers from Norway and Germany have developed an AI algorithm that can predict eye position and movement from MRI images. Their research was published in Nature.
Eye movements are typically recorded with eye-tracking sensory technology in which infrared light is projected into the retina, and then for measurement. Due to strong magnetic fields in MRI, such trackers, or cameras, need to be MRI-compatible. Existing cameras are expensive and are cumbersome to use. As such, until now, eye-tracking has not been a widespread part of MRI examinations.
This may now change, however, thanks to a new open-source ‘DeepMReye’ software, available for download on github. To create the software, the researchers trained a neural network with their own and publicly available data of both eye-tracking and MRI images.
Their software can now predict whether eyes are open or closed and even track eye movement when eyes are closed- without using eye trackers during MRI. Decoding eye movements, say the researchers, can help explain network-wide brain activity as well as those associated with oculomotor function.
The researchers hope that their software will make eye tracking more widely available for MRI research. They say, for example, that the software could be used by sleep researchers to study eye movements in different sleep stages. The software also has the potential for use by clinicians to diagnose disorders that manifest in changes in eye movement patterns.
The researchers add that traditional eye-tracking cameras are rarely used for blind patients due to difficulty in proper calibration. However, they say that DeepMReye overcomes this challenge as the AI ‘can be calibrated with the help of healthy subjects and then be applied in examinations of blind patients.’
Sources: Nature, github, Neuroscience News
