Using two-dimensional oculomotor modeling to analyze smooth pursuit performance

During smooth pursuit, the oculomotor system needs to multi-task: eye velocity is adjusted to keep the moving stimulus on the fovea; in case of target blanking or loss of foveation, saccades are performed to catch up with the target. Mathematical models can be used to implement a quantitative description of underlying processes. Thus, models can provide insight by identifying parameters related to the physiological mechanisms (like pursuit gain or retinal delay).Here, we adapted a one-dimensional oculomotor model (Marti et al. 2008) to a two-dimensional pursuit task with target blanking (see figure 1). While the model was previously used to successfully model oculomotor behaviour, we fitted the model for the first time to individual data recorded from a normative population (Rao et al. 2021). We used Bayesian optimization to minimize the error between experimental and model slow-phase eye velocity depending on a set of selected model parameters.We analyzed the dependence of fitting error on data analysis and extracted empirical ranges of nine parameters and their variability. As expected from formal model analysis, we found a reciprocal relation between brainstem integrator time constant and Purkinje cell gain factor. The resulting models had mean fitting error of 5.49±0.61°/s across participants (target peak velocity 20.42°/s), showing the suitability of the model for individual data analysis.By estimating baseline ranges for normative system parameters, we can identify deviations in non-normative participants. Thus, we can locate whether and where damage occurred, aiding in determining optimal therapeutic strategies.Figure 1: Model Outline