Adaptation in multisensory neurons


Kuang, Xutao (2008) Adaptation in multisensory neurons. University of Southampton, School of Electronics and Computer Science, Doctoral Thesis , 201pp.

Download

[img] PDF
Download (2581Kb)

Description/Abstract

The most studied region in the mammalian brain for multisensory integration is the deep superior colliculus (DSC). Neurophysiological experiments have revealed many response properties of DSC neurons, such as cross-modal enhancement (CME) and sub-additive/additive/super-additive op-
erational modes. CME occurs when the response of a multisensory neuron to stimulation in one sensory modality is enhanced, often non-linearly, by temporally and spatially coincident stimulation of a second sensory modality. Response enhancement is frequently larger for weaker input stimuli than for stronger stimuli, a phenomenon known as inverse e®ectiveness. It is believed that a non-linear, saturating response function may underlie CME associated with inverse effectiveness. We explore this idea in more detail, showing that apart from CME, many other response properties of DSC neurons, including the different dynamic ranges of responses to unimodal and multimodal stimuli and the diverse operational modes, also emerge as a direct consequence of a saturating response function such as a sigmoidal function.

We then consider the question of how the exact form of a candidate, saturating sigmoidal function could be determined in a DSC neuron. In particular, we suggest that adaptation may determine its exact form. Adaptation to input statistics is a ubiquitous property of sensory neurons. Defining the operating point as the output probability density function, we argue that a neuron maintains an invariant operating point by adapting to the lowest-order moments of the input probability distribution. Based on this notion, we propose a novel adaptation rule that permits
unisensory neurons to adapt to the lowest-order statistics of their inputs, and then extend this rule to allow adaptation in multisensory neurons, of which DSC neurons are an example. Adaptation in DSC neurons is expected to change the responses of a neuron to a fixed, probe or test stimulus. Such a neuron would therefore exhibit different CME when presented with the same stimulus drawn from different statistical ensembles. We demonstrate that, for suitable selections
of test stimuli, adaptation to an increase in the mean, the variance or the correlation coefficient induce consistent changes in CME. By virtue of the robustness of the results, the underlying
adaptation notion can be tested in neurophysiological experiments. Finally, it is known that descending cortical projections from the anterior ectosylvian sulcus and the rostral aspect of the lateral suprasylvian sulcus are indispensable for DSC neurons to exhibit CME. The structure of our proposed adaptation rule for multisensory neurons therefore permits us to speculate that the descending cortical inputs to multisensory DSC neurons facilitate the computation of the correlation coefficient between different sensory channels' activities.

Item Type: Thesis (Doctoral)
Subjects: Q Science > QA Mathematics > QA75 Electronic computers. Computer science
Q Science > QP Physiology
Divisions: University Structure - Pre August 2011 > School of Electronics and Computer Science > Intelligence, Agents, Multimedia
ePrint ID: 65419
Date Deposited: 25 Mar 2009
Last Modified: 27 Mar 2014 18:46
URI: http://eprints.soton.ac.uk/id/eprint/65419

Actions (login required)

View Item View Item