Sina M Kling¹, Uriel Lascombes¹, Matthias Nau², Guillaume S Masson¹, Martin Szinte¹; ¹Institut de Neurosciences de la Timone, CNRS, Aix-Marseille Université, Marseille, France, ²Vrije Universiteit Amsterdam, Amsterdam, The Netherlands

Eye movements, even with eyes closed, provide valuable insights into human cognition and are a critical variable in numerous functional magnetic resonance imaging (fMRI) studies. Here, we track eye movements while the eyes are closed using DeepMReye, a deep learning framework for camera-less eye tracking in fMRI. We designed an experiment where participants moved their gaze towards a sequence of known positions, with both eyes open and closed, under conditions with and without visual input, while fMRI data were being acquired. While DeepMReye was initially trained on fMRI data from classical eyes-open tasks, the network could successfully generalize to decode gaze position during eyes-closed periods. Furthermore, model performance improved when the network was specifically trained to generalize across conditions, including eyes-closed and varying visual input. These findings demonstrate that reliable eye movement monitoring during eyes-closed periods is feasible in fMRI, enabling a more effective integration of eye tracking in fMRI research and therefore advancing our understanding of human cognition.

Acknowledgements: This research was supported by an ANR JCJC and a Fyssen Foundation grant to MS.