Effective connectivity in cortical networks underlying visual motion processing and ocular-motor control

Introduction: Human ocular motor functions and visual motion processing have mostly been assumed to rely on segregated networks based on recordings in non-human primates and conventional fMRI studies. Voluntary optokinetic ("look") nystagmus in the horizontal plane, however, contains all essential forms of eye movements whilst at the same time representing a visual motion-tracking task. Aim of the presented approach was to derive an integrative model (“common trunk”) for ocular motor and visual motion networks by means of dynamic causal modeling (DCM).

Methods: The differenzial effects of a horizontal optokinetic stimulation versus a flashing visual control stimulus were studied in 21 right-handed healthy volunteers in a 3 T scanner with a 32-channel head coil. The time course for eight regional responses were extracted for all subjects through a general linear model analysis (p < 0.05 FDR corrected at cluster level) after standard SPM preprocessing. Hence we estimated 2424 different, anatomically possible models for each subject in SPM8 with the DCM 10 framework and performed a random as well as a fixed effects bayesian model selection (BMS) to identify the most likely generative model of our data.

Results: Model evidence in fixed and random BMS converged on the same network featuring a bidirectional connectivity of the frontal (FEF) and parietal eye fields (PEF) (Fig. 1). The network contained a signal loop through a backward connection of the frontal eye fields towards MT/V5. BMS also showed that all driving inputs entered our network from subcortical structures in the respective primary visual cortex V1 and not directly into MT/V5 or even the frontal eye fields. Direct input from V1 was only observed for MT/V5 and the frontal eye fields. The direction of horizontal OKN was reflected in a significant modulation of the forward connection from MT/V5 and the FEF respectively towards the parietal eye fields within each hemisphere. Hemispheric crosstalk (i.e., transcallosal connectivity) between the paired regions (V1, MT/V5, PEF or FEF) was not found to be of relevance within the context of our model.

Fig. 1: Illustration of the winning model. Black arrows indicate the signal direction of significant connections for each region with respect to each hemisphere (l = left, r = right as a prefix). The green arrows point out the entrance of the driving input for left and right OKN. Significant modulations of connections after a change in direction of horizontal OKN are marked by red arrows. Please note the absence of any interhemispheric connections in our network. Each hemisphere seems to process the visual motion paradigm in an autonomic manner.

Conclusions: Our study shows that a unifying cerebral network for visual motion and ocular motor processing can be modelled from fMRI data by means of DCM. The model with the highest posterior indicated a looped convergence of ocular motor and motion information in MT/V5 and the parietal eye fields within the context of our task. The primary visual cortex V1 seems to represent the main cortical gateway for all relevant signals. The future step of building a more encomprising model of visual motion processing and eye movement is now to additionally include secondary areas such as V6, pre-SMA and the supplementary eye fields.