Biomedical subjects

Eran Stark

Publications and source records attributed to Eran Stark.

5 recordsLinked to original sources

PubMed2006

Spike sorting: Bayesian clustering of non-stationary data.

Spike sorting involves clustering spikes recorded by a micro-electrode according to the source neurons. It is a complicated task, which requires much human labor, in part due to the non-stationary nature of the data. We propose to automate the clustering process in a Bayesian framework, with the source neurons modeled as a non-stationary mixture-of-Gaussians. At a first search stage, the data are divided into short time frames, and candidate descriptions of the data as mixtures-of-Gaussians are computed for each frame separately. At a second stage, transition probabilities between candidate mixtures are computed, and a globally optimal clustering solution is found as the maximum-a-posteriori solution of the resulting probabilistic model. The transition probabilities are computed using local stationarity assumptions, and are based on a Gaussian version of the Jensen-Shannon divergence. We employ synthetically generated spike data to illustrate the method and show that it outperforms other spike sorting methods in a non-stationary scenario. We then use real spike data and find high agreement of the method with expert human sorters in two modes of operation: a fully unsupervised and a semi-supervised mode. Thus, this method differs from other methods in two aspects: its ability to account for non-stationary data, and its close to human performance.

Algorithms↗

PubMed2005

Partial cross-correlation analysis resolves ambiguity in the encoding of multiple movement features.

A classical question in neuroscience is which features of a stimulus or of an action are represented in brain activity. When several features are interdependent either at a given point in time or at distinct points in time, neural activity related to one feature appears to be correlated with other features. Thus techniques that simultaneously consider multiple features cannot account for delayed interdependencies between features. The result is an ambiguity with respect to the encoded features. Here, we resolve this ambiguity by applying a novel statistical method based on partial cross-correlations. The method yields estimates of linear correlations between neural activity and a given feature that are not affected by linear correlations with other features at multiple time delays. The method also provides a graphical output measured on a scale that allows for comparisons between different features, neurons, and experiments. We use real movement data and neural activity simulated according to a wide range of tuning models to illustrate the method. When applied to real neural activity, the procedure yields results that indicate which of the considered features the neural activity is related to and at what time delays.

Action Potentials↗

PubMed2005

Applying resampling methods to neurophysiological data.

Standard statistical techniques do not always provide answers to complex physiological questions because often there are no parametric or non-parametric distributions on which significance can be estimated. Resampling methods provide a battery of tests that can be used in such circumstances. In the past few years these methods have been explored theoretically and are now employed frequently. In this paper we describe a unified framework for the use of such methods in the context of neurophysiological data analysis. We construct specific tests for placing confidence limits on estimates of mutual information and on parameters of circular data, and we present procedures for testing hypotheses on circular and on partitioned data. These tests are explained in detail and illustrated with real data from experiments with behaving monkeys.

Algorithms↗

PubMed2003

Neuronal activity in motor cortical areas reflects the sequential context of movement.

Natural actions can be described as chains of simple elements, whereas individual motion elements are readily concatenated to generate countless movement sequences. Sequence-specific neurons have been described extensively, suggesting that the motor system may implement temporally complex motions by using such neurons to recruit lower-level movement neurons modularly. Here, we set out to investigate whether activity of movement-related neurons is independent of the sequential context of the motion. Two monkeys were trained to perform linear arm movements either individually or as components of double-segment motions. However, comparison of neuronal activity between these conditions is delicate because subtle kinematic variations generally occur within different contexts. We therefore used extensive procedures to identify the contribution of variations in motor execution to differences in neuronal activity. Yet, even after application of these procedures we find that neuronal activity in the motor cortex (PMd and M1) associated with a given motion segment differs between the two contexts. These differences appear during preparation and become even more prominent during motion execution. Interestingly, despite context-related differences on the single-neuron level, the population as a whole still allows a reliable readout of movement direction regardless of the sequential context. Thus the direction of a movement and the sequential context in which it is embedded may be simultaneously and reliably encoded by neurons in the motor cortex.

Action Potentials↗

PubMed2003

Dynamical organization of directional tuning in the primate premotor and primary motor cortex.

Although previous studies have shown that activity of neurons in the motor cortex is related to various movement parameters, including the direction of movement, the spatial pattern by which these parameters are represented is still unresolved. The current work was designed to study the pattern of representation of the preferred direction (PD) of hand movement over the cortical surface. By studying pairwise PD differences, and by applying a novel implementation of the circular variance during preparation and movement periods in the context of a center-out task, we demonstrate a nonrandom distribution of PDs over the premotor and motor cortical surface of two monkeys. Our analysis shows that, whereas PDs of units recorded by nonadjacent electrodes are not more similar than expected by chance, PDs of units recorded by adjacent electrodes are. PDs of units recorded by a single electrode display the greatest similarity. Comparison of PD distributions during preparation and movement reveals that PDs of nearby units tend to be more similar during the preparation period. However, even for pairs of units recorded by a single electrode, the mean PD difference is typically large (45 degrees and 75 degrees during preparation and movement, respectively), so that a strictly modular representation of hand movement direction over the cortical surface is not supported by our data.

Animals↗