Synergistic Coding by Cortical Neural Ensembles

Functional connectivity dynamics among cortical neurons: a dependence analysis

This paper quantifies and comparatively validates functional connectivity between neurons by measuring the statistical dependence between their firing rates. Based on statistical analysis of the pairwise functional connectivity, we estimate, exclusively from neural data, the neural assembly functional connectivity given a behavior task, which provides a quantifiable representation of the dynamic nature during the behavioral task. Because of the time scale of behavior (100-1000 ms), a statistical method that yields robust estimators for this small sample size is desirable. In this work, the temporal resolutions of four estimators of functional connectivity are compared on both simulated data and real neural ensemble recordings. The comparison highlights how the properties and assumptions of statistical-based and phase-based metrics affect the interpretation of connectivity. Simulation results show that mean square contingency (MSC) and mutual information (MI) create more robust quantification of functional connectivity under identical conditions than cross correlation (CC) and phase synchronization (PhS) when the sample size is 1 s. The results of the simulated analysis are extended to real neuronal recordings to assess the functional connectivity in monkey's cortex corresponding to three movement states in a food reaching task and construct the assembly graph given a movement state and the activation degree of a state-related assembly over time using the statistical test exclusively from neural data dependencies. The activation degree of a given state-related assembly reaches the peak repeatedly when the specific movement states occur, which also reveals the network of interactions among the neurons are key for the operation of a specific behavior.

Neural feedback for instantaneous spatiotemporal modulation of afferent pathways in bi-directional brain-machine interfaces

In bi-directional brain-machine interfaces (BMIs), precisely controlling the delivery of microstimulation, both in space and in time, is critical to continuously modulate the neural activity patterns that carry information about the state of the brain-actuated device to sensory areas in the brain. In this paper, we investigate the use of neural feedback to control the spatiotemporal firing patterns of neural ensembles in a model of the thalamocortical pathway. Control of pyramidal (PY) cells in the primary somatosensory cortex (S1) is achieved based on microstimulation of thalamic relay cells through multiple-input multiple-output (MIMO) feedback controllers. This closed loop feedback control mechanism is achieved by simultaneously varying the stimulation parameters across multiple stimulation electrodes in the thalamic circuit based on continuous monitoring of the difference between reference patterns and the evoked responses of the cortical PY cells. We demonstrate that it is feasible to achieve a desired level of performance by controlling the firing activity pattern of a few "key" neural elements in the network. Our results suggest that neural feedback could be an effective method to facilitate the delivery of information to the cortex to substitute lost sensory inputs in cortically controlled BMIs.

The Ising decoder: reading out the activity of large neural ensembles

The Ising model has recently received much attention for the statistical description of neural spike train data. In this paper, we propose and demonstrate its use for building decoders capable of predicting, on a millisecond timescale, the stimulus represented by a pattern of neural activity. After fitting to a training dataset, the Ising decoder can be applied "online" for instantaneous decoding of test data. While such models can be fit exactly using Boltzmann learning, this approach rapidly becomes computationally intractable as neural ensemble size increases. We show that several approaches, including the Thouless-Anderson-Palmer (TAP) mean field approach from statistical physics, and the recently developed Minimum Probability Flow Learning (MPFL) algorithm, can be used for rapid inference of model parameters in large-scale neural ensembles. Use of the Ising model for decoding, unlike other problems such as functional connectivity estimation, requires estimation of the partition function. As this involves summation over all possible responses, this step can be limiting. Mean field approaches avoid this problem by providing an analytical expression for the partition function. We demonstrate these decoding techniques by applying them to simulated neural ensemble responses from a mouse visual cortex model, finding an improvement in decoder performance for a model with heterogeneous as opposed to homogeneous neural tuning and response properties. Our results demonstrate the practicality of using the Ising model to read out, or decode, spatial patterns of activity comprised of many hundreds of neurons.