Eye velocity gain fields for visuo- motor coordinate transformations

’Gain-field-like’ tuning behavior is characterized by a modulation
of the neuronal response depending on a certain variable, without
changing the actual receptive field characteristics in relation to
another variable. Eye position gain fields were first observed in
area 7a of the posterior parietal cortex (PPC), where visually
responsive neurons are modulated by ocular position. Analysis of
artificial neural networks has shown that this type of tuning
function might comprise the neuronal substrate for coordinate
transformations. In this work, neuronal activity in the dorsal
medial superior temporal area (MSTd) has been analyzed with an
focus on it’s involvement in oculomotor control. MSTd is part of
the extrastriate visual cortex and located in the PPC. Lesion
studies suggested a participation of this cortical area in the
control of eye movements. Inactivation of MSTd severely impairs the
optokinetic response (OKR), which is an reflex-like kind of eye
movement that compensates for motion of the whole visual scene.
Using a novel, information-theory based approach for neuronal data
analysis, we were able to identify those visual and eye movement
related signals which were most correlated to the mean rate of
spiking activity in MSTd neurons during optokinetic stimulation. In
a majority of neurons firing rate was non-linearly related to a
combination of retinal image velocity and eye velocity. The
observed neuronal latency relative to these signals is in line with
a system-level model of OKR, where an efference copy of the motor
command signal is used to generate an internal estimate of the
head-centered stimulus velocity signal. Tuning functions were
obtained by using a probabilistic approach. In most MSTd neurons
these functions exhibited gain-field-like shapes, with eye velocity
modulating the visual response in a multiplicative manner.
Population analysis revealed a large diversity of tuning forms
including asymmetric and non-separable functions. The distribution
of gain fields was almost identical to the predictions from a
neural network model trained to perform the summation of image and
eye velocity. These findings therefore strongly support the
hypothesis of MSTd’s participation in the OKR control system by
implementing the transformation from retinal image velocity to an
estimate of stimulus velocity. In this sense, eye velocity gain
fields constitute an intermediate step in transforming the
eye-centered to a head-centered visual motion signal.Another aspect
that was addressed in this work was the comparison of the
irregularity of MSTd spiking activity during optokinetic response
with the behavior during pure visual stimulation. The goal of this
study was an evaluation of potential neuronal mechanisms underlying
the observed gain field behavior. We found that both inter- and
intra-trial variability were decreased with increasing retinal
image velocity, but increased with eye velocity. This observation
argues against a symmetrical integration of driving and modulating
inputs. Instead, we propose an architecture where multiplicative
gain modulation is achieved by simultaneous increase of excitatory
and inhibitory background synaptic input. A conductance-based
single-compartment model neuron was able to reproduce realistic
gain modulation and the observed stimulus-dependence of neural
variability, at the same time. In summary, this work leads to
improved knowledge about MSTd’s role in visuomotor transformation
by analyzing various functional and mechanistic aspects of eye
velocity gain fields on a systems-, network-, and neuronal level.