| || |
|Title:||A NETWORK OF SPIKING NEURONS PREDICTS EYE MOVEMENT DECISIONS IN A VISUAL DISCRIMINATION TASK UNDER IMPAIRED NMDA FUNCTION|
| || |
|Authors/Affiliations:||1 Dominic Standage*; 1 Martin Pare; |
1 Queen's University, Kingston, ON, Canada
| || |
|Content:||Objectives: Decision-related neural activity in higher association cortices has been the subject of intense research interest in recent years. Building on the model of Wang and colleagues (Compte et al., Cerebral Cortex, 2000; Wang, J. Neurosci., 2002; Wong and Wang, J. Neurosci. 2006), we investigate the role of NMDA receptors (NMDAR's) in the recurrent circuitry of these putative decision circuits. Using a biophysically-based network model, we simulate our own laboratory experiment (see Kalwarowsky et al. in these proceedings) where following administration of the NMDA antagonist Ketamine, monkeys are required to foveate a target in an array of distractors. Experimental results show slower, more accurate eye-movement decisions following Ketamine injection than in control trials.|
Materials and Methods: We use a fully-connected, recurrent network of spiking neurons, where all parameter values are guided by experimental measurements including a 4-to-1 ratio of pyramidal cells to interneurons. Recurrent activity from pyramidal cells is mediated by AMPA and NMDA conductances, and from interneurons by GABA conductances. The strength of pyramidal-to-pyramidal connections is a Gaussian function of the distance between neurons. All other connections are unstructured. This connectivity imposes local cooperation between pyramidal cells and global competition via interneurons. Visual stimuli are simulated by Poisson spike trains, where spike rates are drawn from a normal distribution and the mean corresponds to the centre of a Gaussian-shaped receptive field. Afferent background activity is also mediated by AMPA receptors (AMPAR's), causing baseline spiking of 1 or 2 Hz. We simulate the effect of Ketamine by reducing the strength of NMDA conductances.
Results: Simulations are consistent with our experimental results: reduction in the strength of NMDA conductances causes a slightly slower network dynamic, yielding slower (simulated) decision times. At first glance, our results appear to conflict with predictions of Wang's (2002) model, where increasing the strength of NMDAR's (within a restricted range) slowed down the network, and with Wong and Wang's (2006) reduced model, where an increase in the ratio of AMPAR's to NMDAR's lead to faster, less accurate decisions. However, there are a number of crucial differences between our simulations and theirs. Firstly, Wang's (2002) finding was obtained in a parameter regime providing self-sustaining recurrent activity. We tune our network to not support such memory states, and indeed, Wong and Wang (2006) demonstrated very slow decision times in a model without self-sustaining states. Furthermore, Wong and Wang kept the total strength of AMPAR's and NMDAR's constant when manipulating their ratio. Because Ketamine has no effect on AMPAR's, we only reduce the strength of NMDAR's.
Conclusions: A recurrent network of spiking neurons predicts eye movement decisions under impaired NMDA function. Our presentation includes (i) simulations of our experimental manipulation of target-distractor similarity, replicating our psychometric data, and (ii) manipulation of the ratio of afferent synaptic strength to recurrent strength, addressing the source of delay activity in decision-related association cortices.
| || |