Machine Learning Techniques to extract eye movement trajectories from MRI data

Eye movement artefacts in MRI scans remain unresolved, hampering ophthalmological examinations. To eliminate these artefacts, a recent technique relies on eye tracking using an additional device installed in the scanner. However, this technique can only track one eye at a time and is costly in terms of equipment (which must be compatible with MRI), installation time, and qualified personnel.

The MREye-Track project aims to introduce a new approach that will address the limitations of current Magnetic Resonance Eye Imaging practice by employing machine learning techniques to raw MRI data (frequency domain) for the automatic identification of i) acquisitions affected by eye movements and ii) gaze direction, to reconstruct a high-quality anatomical image of the eye without additional equipment. By enabling the analysis of both eyes from the same scan and minimizing the need for hospital resources and examination time, MREye-Track will markedly improve the accessibility of ocular MRI.