Limits...
Fast, scalable, Bayesian spike identification for multi-electrode arrays.

Prentice JS, Homann J, Simmons KD, Tkačik G, Balasubramanian V, Nelson PC - PLoS ONE (2011)

Bottom Line: Our method can distinguish large numbers of distinct neural units, even when spikes overlap, and accounts for intrinsic variability of spikes from each unit.Human interaction plays a key role in our method; but effort is minimized and streamlined via a graphical interface.We illustrate our method on data from guinea pig retinal ganglion cells and document its performance on simulated data consisting of spikes added to experimentally measured background noise.

View Article: PubMed Central - PubMed

Affiliation: Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, Pennsylvania, United States of America. jprentic@sas.upenn.edu

ABSTRACT
We present an algorithm to identify individual neural spikes observed on high-density multi-electrode arrays (MEAs). Our method can distinguish large numbers of distinct neural units, even when spikes overlap, and accounts for intrinsic variability of spikes from each unit. As MEAs grow larger, it is important to find spike-identification methods that are scalable, that is, the computational cost of spike fitting should scale well with the number of units observed. Our algorithm accomplishes this goal, and is fast, because it exploits the spatial locality of each unit and the basic biophysics of extracellular signal propagation. Human interaction plays a key role in our method; but effort is minimized and streamlined via a graphical interface. We illustrate our method on data from guinea pig retinal ganglion cells and document its performance on simulated data consisting of spikes added to experimentally measured background noise. We present several tests demonstrating that the algorithm is highly accurate: it exhibits low error rates on fits to synthetic data, low refractory violation rates, good receptive field coverage, and consistency across users.

Show MeSH

Related in: MedlinePlus

Visualization of data; flowchart of our method.(A)  firing events cluster well in terms of their barycenter (voltage weighted average spatial location) and absolute peak voltage (see text), despite wide amplitude dispersion in some groups; each combination of color and marker size corresponds to one spiking unit identified by the clustering method developed in the text. Grey points were unassigned to any cluster. A total of 107 clusters are marked. (B) Events cluster poorly when projected onto the three principal features uncovered by principal component analysis (PCA). Coloring as in (A). (C) Schematic of our spike sorting method. Dashed lines involve a small subset of a full dataset. The backwards arrow describes the introduction of new spike templates found after the first pass of fitting; a total of two passes are performed. (D) The optics algorithm orders all firing events into a linear sequence based on a distance measure (see text). Events are lined up in this order (x-axis), and represented in terms of the 960 voltage samples recorded by all the electrodes during a 3.2 ms firing event (y-axis; from top to bottom, 32 consecutive time samples from one channel, then 32 time samples from the second channel, and so on). The human operator highlights bands of events (typically very clear to an observer) that appear to constitute a single cluster; one such band is shown. Later automated diagnostics refine and check these assessments.
© Copyright Policy
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3140468&req=5

pone-0019884-g002: Visualization of data; flowchart of our method.(A) firing events cluster well in terms of their barycenter (voltage weighted average spatial location) and absolute peak voltage (see text), despite wide amplitude dispersion in some groups; each combination of color and marker size corresponds to one spiking unit identified by the clustering method developed in the text. Grey points were unassigned to any cluster. A total of 107 clusters are marked. (B) Events cluster poorly when projected onto the three principal features uncovered by principal component analysis (PCA). Coloring as in (A). (C) Schematic of our spike sorting method. Dashed lines involve a small subset of a full dataset. The backwards arrow describes the introduction of new spike templates found after the first pass of fitting; a total of two passes are performed. (D) The optics algorithm orders all firing events into a linear sequence based on a distance measure (see text). Events are lined up in this order (x-axis), and represented in terms of the 960 voltage samples recorded by all the electrodes during a 3.2 ms firing event (y-axis; from top to bottom, 32 consecutive time samples from one channel, then 32 time samples from the second channel, and so on). The human operator highlights bands of events (typically very clear to an observer) that appear to constitute a single cluster; one such band is shown. Later automated diagnostics refine and check these assessments.

Mentions: We first attempted a “geographical clustering”: from each event we found the minimum of the potential on each channel and the channel containing the deepest minimum (“leader channel”). We then used the absolute values of the minima as weights in a weighted average of the locations of the electrodes neighboring the leader channel. This weighted average gave a particularly salient two-dimensional feature, the event's barycenter:The sums run over the electrodes neighboring the leader, and is the location of the neighbor electrode. We then augmented with a third feature, , minus the absolute peak potential. These three features gave a scatterplot that clearly showed many well-separated clusters (Fig. 2A), without any need to deduce the “best” features by principal component analysis (Fig. 2B).


Fast, scalable, Bayesian spike identification for multi-electrode arrays.

Prentice JS, Homann J, Simmons KD, Tkačik G, Balasubramanian V, Nelson PC - PLoS ONE (2011)

Visualization of data; flowchart of our method.(A)  firing events cluster well in terms of their barycenter (voltage weighted average spatial location) and absolute peak voltage (see text), despite wide amplitude dispersion in some groups; each combination of color and marker size corresponds to one spiking unit identified by the clustering method developed in the text. Grey points were unassigned to any cluster. A total of 107 clusters are marked. (B) Events cluster poorly when projected onto the three principal features uncovered by principal component analysis (PCA). Coloring as in (A). (C) Schematic of our spike sorting method. Dashed lines involve a small subset of a full dataset. The backwards arrow describes the introduction of new spike templates found after the first pass of fitting; a total of two passes are performed. (D) The optics algorithm orders all firing events into a linear sequence based on a distance measure (see text). Events are lined up in this order (x-axis), and represented in terms of the 960 voltage samples recorded by all the electrodes during a 3.2 ms firing event (y-axis; from top to bottom, 32 consecutive time samples from one channel, then 32 time samples from the second channel, and so on). The human operator highlights bands of events (typically very clear to an observer) that appear to constitute a single cluster; one such band is shown. Later automated diagnostics refine and check these assessments.
© Copyright Policy
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3140468&req=5

pone-0019884-g002: Visualization of data; flowchart of our method.(A) firing events cluster well in terms of their barycenter (voltage weighted average spatial location) and absolute peak voltage (see text), despite wide amplitude dispersion in some groups; each combination of color and marker size corresponds to one spiking unit identified by the clustering method developed in the text. Grey points were unassigned to any cluster. A total of 107 clusters are marked. (B) Events cluster poorly when projected onto the three principal features uncovered by principal component analysis (PCA). Coloring as in (A). (C) Schematic of our spike sorting method. Dashed lines involve a small subset of a full dataset. The backwards arrow describes the introduction of new spike templates found after the first pass of fitting; a total of two passes are performed. (D) The optics algorithm orders all firing events into a linear sequence based on a distance measure (see text). Events are lined up in this order (x-axis), and represented in terms of the 960 voltage samples recorded by all the electrodes during a 3.2 ms firing event (y-axis; from top to bottom, 32 consecutive time samples from one channel, then 32 time samples from the second channel, and so on). The human operator highlights bands of events (typically very clear to an observer) that appear to constitute a single cluster; one such band is shown. Later automated diagnostics refine and check these assessments.
Mentions: We first attempted a “geographical clustering”: from each event we found the minimum of the potential on each channel and the channel containing the deepest minimum (“leader channel”). We then used the absolute values of the minima as weights in a weighted average of the locations of the electrodes neighboring the leader channel. This weighted average gave a particularly salient two-dimensional feature, the event's barycenter:The sums run over the electrodes neighboring the leader, and is the location of the neighbor electrode. We then augmented with a third feature, , minus the absolute peak potential. These three features gave a scatterplot that clearly showed many well-separated clusters (Fig. 2A), without any need to deduce the “best” features by principal component analysis (Fig. 2B).

Bottom Line: Our method can distinguish large numbers of distinct neural units, even when spikes overlap, and accounts for intrinsic variability of spikes from each unit.Human interaction plays a key role in our method; but effort is minimized and streamlined via a graphical interface.We illustrate our method on data from guinea pig retinal ganglion cells and document its performance on simulated data consisting of spikes added to experimentally measured background noise.

View Article: PubMed Central - PubMed

Affiliation: Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, Pennsylvania, United States of America. jprentic@sas.upenn.edu

ABSTRACT
We present an algorithm to identify individual neural spikes observed on high-density multi-electrode arrays (MEAs). Our method can distinguish large numbers of distinct neural units, even when spikes overlap, and accounts for intrinsic variability of spikes from each unit. As MEAs grow larger, it is important to find spike-identification methods that are scalable, that is, the computational cost of spike fitting should scale well with the number of units observed. Our algorithm accomplishes this goal, and is fast, because it exploits the spatial locality of each unit and the basic biophysics of extracellular signal propagation. Human interaction plays a key role in our method; but effort is minimized and streamlined via a graphical interface. We illustrate our method on data from guinea pig retinal ganglion cells and document its performance on simulated data consisting of spikes added to experimentally measured background noise. We present several tests demonstrating that the algorithm is highly accurate: it exhibits low error rates on fits to synthetic data, low refractory violation rates, good receptive field coverage, and consistency across users.

Show MeSH
Related in: MedlinePlus