1471-2202-12-S1-P153 1471-2202 Poster presentation <p>A dendrogram approach to the structure of spike trains</p> Houghton Conor houghton@maths.tcd.ie

School of Mathematics, Trinity College Dublin, Dublin, Ireland

BMC Neuroscience <p>Twentieth Annual Computational Neuroscience Meeting: CNS*2011</p> Jean-Marc Fellous and Astrid Prinz Publication of this supplement has been supported by Neuralynx, Springer, INCF and the Bernstein Network Computational Neuroscience Meeting abstracts - A single PDF containing all abstracts in this supplement is available here. http://www.biomedcentral.com/content/pdf/1471-2202-12-S1-info.pdf <p>Twentieth Annual Computational Neuroscience Meeting: CNS*2011</p> Stockholm, Sweden 23-28 July 2011 http://www.cnsorg.org/2011/ 1471-2202 2011 12 Suppl 1 P153 http://www.biomedcentral.com/1471-2202/12/S1/P153 10.1186/1471-2202-12-S1-P153
18 7 2011 2011 Houghton; licensee BioMed Central Ltd. This is an open access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

A novel mathematical description for the temporal structure of spike trains is presented. This works by mapping the spike train to a dendrogram produced by hierarchical clustering. The branch-length structure of the dendrogram is equivalent to the distribution of inter-spike intervals, but morphological descriptions of the dendrogram can be used to quantify other aspects of the temporal structure of spike trains. In this way, it is shown that example sets of spike trains have more structure than is accounted for by the distribution of inter-spike intervals. The goal is to segregate the loose notion of “sparseness” into two different aspects of temporal structure, the distribution of inter-spike intervals and the morphology of the dendrogram constructed during clustering and to provide a tool for analyzing and comparing spike train properties.

Roughly, the distribution of inter-spike intervals fails to describe the temporal granularity caused by the clustering of spikes. The difficulty with quantifying this is in deciding how close two spikes need to be for them to be agglomerated in clusters. The approach taken here is to avoid that choice by performing a hierarchical clustering: the spikes are agglomerated into larger and larger groups as a timescale is increased, this is illustrated in Figure 1. Quantifying morphological properties like the bushiness of the resulting dendrogram then describes the temporal structure of spike trains. In fact, hierarchical clustering is of interest to computer scientists and this provides a number of candidate methods for quantifying the morphology of dendrograms, the one used here was introduced in 1 and scores dendrograms from zero, for the leggiest, to one, for the bushiest.

<p>Figure 1</p>

The spikes are clustered at different time scales to give an agglomerative dendrogram.

The spikes are clustered at different time scales to give an agglomerative dendrogram.

This quantification is applied to an example spike-train data set described in 2 and made available on http://neurodatabase.org. The data consist of single cell recordings from the nucleus of the solitary tract in rat during presentation of taste stimuli. With a clear separation between the stimulus timescale and the temporal scale of spiking, these data are chosen to present a kind of worst-case scenario for the proposal. Nonetheless, when the real data is compared to artificial data produced by shuffling the inter-spike intervals, the Mann–Whitney U test shows that real data has a significantly higher value of the morphological parameter introduced here.

Acknowledgements

Thanks to Patricia Di Lorenzo , Jonathan Victor and http://neurodatabase.org for making the data analyzed here available; Science Foundation Ireland for grant 08/RFP/MTH1280 and the James S McDonnell Foundation for an Understanding Human Cognition Scholar award.

<p>An empirical study of the coeffcients for measuring the structure of hierarchic classification</p> Fionn Murtagh Data Analysis and Informatics III Amsterdam: Elsevier E. Diday et al. 1984 385 393 <p>Taste Response Variability and Temporal Coding in the Nucleus of the Solitary Tract of the Rat</p> Di Lorenzo Patricia Victor Jonathan J Neurophysiology 2003 90 1418 1431 10.1152/jn.00177.2003