Abstract
Purpose :
Video-based eye tracking devices cannot measure translational eye motion and require a calibration process that relies on the participant’s self-report of accurate fixation. To circumvent these limitations, we developed an eye tracking method based on real-time MRI data. We tested this novel method by measuring the participant’s horizontal gaze using real-time MRI and a conventional video-based eye tracker (Eyelink 1000) simultaneously.
Methods :
In order to study the human eye in motion we used a balanced steady-state free precession (bSSFP) MRI sequence, collecting data of a single slice with a temporal resolution of 35ms. We start our analysis with a geometric model of the human eyeball, assuming that the sclera, the outer curvature of the cornea and the inner curvature of the lens take the shape of an ellipsoid. The geometric model is fitted to MRI data by matching the normal vectors of the model to the gradient flow field of the MRI data. We used a high-resolution 3D T2-weighted MRI scan to optimise the model for each eye and then collected dynamical single-slice data of the eye in motion for 2 participants. In particular, we collected binocular horizontal eye movements between targets at -7° and 7° and some blink-related eye movements. The eye movement was estimated as the best projection of the 3D model.
Results :
A linear regression analysis of the horizontal gaze between Eyelink and MRI yielded highly significant results (P<0.001) for both participants and both eyes. The correlation coefficients were 0.99, 0.97, 0.99 and 0.97. Apart from rotation, we also measured translations of up to 1.7mm during blinks.
Conclusions :
We demonstrated a proof of principle that MRI can be used to track the human eye with high temporal resolution using a fully-automated algorithm. We anticipate that our work will lead to a deeper understanding of oculomotor mechanics and will be in particular useful for studying eye movement disorders like nystagmus and strabismus.
This is a 2021 ARVO Annual Meeting abstract.