From brainstem to cortex: Computational models of saccade generation circuitry

Abstract

The brain circuitry of saccadic eye movements, from brainstem to cortex, has been extensively studied during the last 30 years. The wealth of data gathered allowed the conception of numerous computational models. These models proposed descriptions of the putative mechanisms generating this data, and, in turn, made predictions and helped to plan new experiments.

In this article, we review the computational models of the five main brain regions involved in saccade generation: reticular formation saccadic burst generators, superior colliculus, cerebellum, basal ganglia and premotor cortical areas. We present the various topics these models are concerned with: location of the feedback loop, multimodal saccades, long-term adaptation, on the fly trajectory correction, strategy and metrics selection, short-term spatial memory, transformations between retinocentric and craniocentric reference frames, sequence learning, to name the principle ones.

Our objective is to provide a global view of the whole system. Indeed, narrowing too much the modelled areas while trying to explain too much data is a recurrent problem that should be avoided. Moreover, beyond the multiple research topics remaining to be solved locally, questions regarding the operation of the whole structure can now be addressed by building on the existing models.

Introduction

There are three main different types of primate eye movements (slow, fast and vergence movements), that are controlled by partially separate brain structures (Henn, 1993). The slow movements include the vestibulo-ocular reflex, the slow-phase of the optokinetic reflex and the smooth pursuit. The fast movements are the fast-phase of the optokinetic reflex and the saccades. Saccades are used by species (like humans and primates) whose retinas have a central high-resolution region (the fovea) to explore visual scenes by redirecting gaze from one important visual stimulus requiring precise analysis to another. Their speed may reach 1000 $°{\text{s}}^{-1}$ in some primate species.

The mechanics of saccadic eye movements are relatively simple when compared to limb movements, which use multiple joints and operate with varying loads. Saccadic eye movements have therefore been studied for the intrinsic interest of understanding how they are generated, but also as a simple way to more generally study motor and premotor mechanisms in the brain.

Numerous brain regions are involved in the generation of saccades Berthoz, 1996, Moschovakis et al., 1996, from the cortex down to the brainstem (Fig. 1). The closest to the movement execution are the vertical and horizontal saccadic burst generators (SBG), two sets of nuclei of the reticular formation which directly drive the ocular motoneurons (Scudder et al., 2002). Their function is to produce, from eye displacement instructions issued from higher level structures, the commands appropriate to generate saccades with the desired metrics. They are supposed to ensure accuracy by monitoring their own commands through an efferent copy-based feedback.

The superior colliculus (SC) is, with the frontal eye fields (FEF), the main structure sending saccade orders to the SBG (Moschovakis, 1996). The SC is a place of convergence and integration, often designed as the final common path of saccades. It receives projections carrying simple visual, auditive and somatosensory information along with more cognitive signals, where the sensory inputs are affected by attention, motivation and context. The SC drives the orientation of the whole body: it does not control the eye direction with regards to the head, but the gaze direction. Therefore, it activates not only the SBG but also, for example, the neck muscles.

The commands directed from the SC to the SBG are under the influence of adaptive modulations issued from the cerebellum (CBLM). It provides the SBG with additional input during saccades which are interpreted as: (1) a calibration of the system induced by long-term adaptation of the saccadic gain and (2) an on the fly correction of every single saccade, made necessary by the apparent variability of the rest of the saccade generating circuitry (Optican and Robinson, 1980).

The activity of the SC is gated by inhibitory inputs issued from a set of subcortical nuclei called the basal ganglia (BG) Hikosaka and Wurtz, 1983a, Hikosaka and Wurtz, 1983b, Hikosaka and Wurtz, 1983c, Hikosaka and Wurtz, 1983d, Chevalier and Deniau, 1993, Hikosaka et al., 2000. Whereas the cortical areas generate numerous motor orders which are directly sent to the SC, they also project to the BG which is implicated in choosing which orders to execute, by disinhibiting the corresponding subregion of the SC. The role of the BG might however not be restricted to that metric selection (Handel and Glimcher, 1999).

Atop all these structures, many cortical areas are involved in saccade generation: the posterior parietal cortex (PPC), the dorsolateral prefrontal cortex (DLPFC), the anterior cingulate cortex (aCG), the presupplementary, supplementary and frontal eye fields (pre-SEF, SEF and FEF, respectively) (Platt et al., 2004). They provide rich inputs for the SC which allow the selection of targets by cognitive processes influenced by motivational and attentional states, along with the possible use of working memory or sequence learning capabilities (Pierrot-Deseilligny et al., 2003).

Numerous computational models of all these saccade-related brain regions have been proposed in the last 30 years. They helped to understand their operation, functionality and interconnections by proposing computational mechanisms and predictions that could be tested experimentally. However, most of them were restricted to one or a few subparts of the whole circuitry. This may sometimes cause problems when one attempts to replicate with such restricted models experimental results that are indeed generated by another brain structure or by interactions with other structures. The objective of this paper is to review the computational models of saccade-related brain circuits, from brainstem to cortex, in order to propose an ensemble view of the system, of the numerous problems remaining to be solved at each level and of the relationships between each levels.

The computational models of the five categories of brain areas involved in saccade generation are reviewed in the next sections in the following order: reticular formation saccadic burst generators, superior colliculus, cerebellum, basal ganglia and cortex. Each of these sections has its own specific discussion, while global considerations are provided in a final conclusion.