We will be presenting two posters at this years COSYNE conference in Salt Lake City Utah.
Thurs Feb 25 8pm
I-89. Adaptive thalamic gating: A framework of dynamic encoding Clarissa Whitmire, Christian Waiblinger, Cornelius Schwarz, Garrett Stanley, Georgia Institute of Technology
Friday Feb 26 7:30pm
II-42. Closed loop optogenetic control of neural circuits: Tracking dynamic trajectories of neural activity Michael Bolus, Adam Willats, Clarissa Whitmire, Zak Costello, Magnus Egerstedt, Christopher Rozell, Garrett Stanley, W.H. Coulter Dept. BME, GT
Abstracts are below:
I-89. Adaptive thalamic gating: A framework of dynamic encoding Clarissa Whitmire1,2 CLARISSA.WHITMIRE@GATECH.EDU Christian Waiblinger1,2 CHRISTIAN.WAIBLINGER@GATECH.EDU Cornelius Schwarz3 CORNELIUS.SCHWARZ@UNI-TUEBINGEN.DE Garrett Stanley1,2 GARRETT.STANLEY@BME.GATECH.EDU 1Georgia Institute of Technology 2Emory University 3University of Tuebingen It has been posited that the regulation of burst/tonic firing in the thalamus could function as a mechanism for controlling not only how much, but what kind of information is conveyed to downstream cortical targets. Yet how this gating mechanism is adaptively modulated on fast time scales by ongoing sensory inputs in rich sensory environments remains unknown. Using single unit recordings in the rat vibrissa thalamus (VPm), we found that the degree of adaptation modulated thalamic burst/tonic firing as well as the synchronization of bursting across the thalamic population along a continuum for which the extremes facilitate detection or discrimination of sensory inputs. Optogenetic control of thalamic state combined with computational modeling of single neuron dynamics further suggests that this regulation of burst/tonic firing may result from an interplay between adaptive changes in thalamic membrane potential and reduced synaptic drive from inputs to thalamus. Consistent with the view that tonic firing relays detailed information while burst firing signals the presence of a stimulus, parsing trials by burst and tonic responses demonstrated that thalamic bursting facilitated detectability while tonic activity facilitated discriminability from the perspective of an ideal observer. Generalized linear model (GLM) fits of the thalamic activity in different optogenetically manipulated thalamic states revealed clear feature selectivity associated with tonic firing, yet the thalamic bursting activity was not well captured by the standard GLM architecture due to an extreme dependence upon the silence between spiking periods, or spike history, beyond that of the standard GLM. As such, we hypothesize that a more accurate method of burst encoding models will require the tonic firing estimate of feature selectivity combined with a state variable to estimate bursting affinity. Taken together, these results suggest that dynamic burst/tonic thalamic encoding sets the stage for an intricate control strategy upon which cortical computation is built.
II-42. Closed loop optogenetic control of neural circuits: Tracking dynamic trajectories of neural activity Michael Bolus1,2 MBOLUS@BELLSOUTH.NET Adam Willats3,2 AWILLATS3@GATECH.EDU Clarissa Whitmire3,2 CLARISSA.WHITMIRE@GATECH.EDU Zak Costello3 ZAK.COSTELLO@GATECH.EDU Magnus Egerstedt3 MAGNUS.EGERSTEDT@ECE.GATECH.EDU Christopher Rozell3 CROZELL@GATECH.EDU Garrett Stanley3,2 GARRETT.STANLEY@BME.GATECH.EDU 1W.H. Coulter Dept. BME, GT 2Emory University 3Georgia Institute of Technology Previously we demonstrated using closed loop optogenetic control to clamp the firing rate of single neurons to static targets. Clamping firing rates to fixed levels has many potential applications as a systems-level analog of Hodgkin and Huxley’s voltage clamp. However, time-varying activities in the brain, such as mode-switching and brain state oscillations, are thought to be critical in mediating behavior and perception, which motivates developing a tool to replicate and manipulate these signals. Here we develop a framework for tracking dynamic trajectories, using optogenetic control of single unit thalamic firing activity in-vivo. Specifically, we utilize the thalamocortical circuit in the rat vibrissa pathway as a model system, closing the loop around the spiking activity of single units expressing channelrhodopsin-2 (ChR2) in somatosensory thalamus. Integral to the framework is the implementation of an observer designed to estimate the latent firing rate from spiking activity, providing feedback for control. We take a first order approach to observing firing rate using a fixed bandwidth exponential filter as well as a more sophisticated adaptive approach. Experimentally and computationally we find controller performance is highly sensitive to observer design. When the time constant of the exponential filter (tau) is tuned according to the target frequency, control of some dynamic targets can be achieved but at the cost of increasingly noisy estimates of firing rate. To ameliorate this tradeoff in controller performance, we have taken an adaptive point process filter (aPPF) approach, whereby the filter used to estimate firing rate is modulated according to spiking activity itself. Using an aPPF improves not only the ability to track dynamic firing rate trajectories, but also the fidelity of the estimate. Taken together, these developments represent a fundamental building block for control of neural activity which we are extending to larger-scale problems of multi-unit-, population-, and systems-level control
