SES Toolbox webpage


Home               FAQ               Download               Discussion group               Contact us

 
Welcome to the SES project homepage!
 

SES stands for Stochastic Event Synchrony, and is a family of similarity measure for point processes. SES can be applied to one-dimensional (e.g., spike trains) and multi-dimensional point processes (e.g., sparse time-frequency representations of electrophysiological signals such as matching-pursuit representations, chirplets, Hilbert-Huang transforms, and bump models).

SES tries to align events in the point processes; the better the alignment, the more similar the point processes are considered to be. More precisely, the similarity is quantified by the following parameters: time delay, variance of the timing jitter, fraction of "non-aligned'' events, and average similarity of the aligned events.

The SES measures may be viewed as extensions of the cost-based metrics of Victor et al. SES gives a statistical interpretation of those metrics, and therefore, it is able to automatically infer the unit costs, in contrast to the cost-based metrics. The latter are so far only applicable to one-dimensional point processes, whereas SES is applicable to multi-dimensional point processes as well.

We have used SES to:

·         quantify the firing reliability of Morris-Lecar neurons,

·         detect loss of EEG synchrony in Mild Cognitive Impairment patients,

·         analyze steady-state visually evoked potentials and EEG responses to auditory stimuli,

·         investigate the causal relation between morphological and molecular signaling events in cell migration.

 


Fig.1. Generative statistical model that describes SES for two given spike trains x and x'.

 

Literature

Theory and Algorithms

SES is described in detail in the following two papers:

A compact conference version of the above two papers:

The first paper (Part I) describes the SES similarity measures for pairs of one-dimensional point processes; it is used to quantify the firing reliability of Morris-Lecar neurons (type I and II).

The second paper (Part II) considers the extension to pairs of multi-dimensional point processes. As an illustration, SES is applied to sparse time-frequency representations of electroencephalograms (EEG), referred to as "bump models". We apply SES to detect loss in EEG synchrony in Mild Cognitive Impairment (MCI) patients.

A third paper is on its way, describing the extension from pairs of point processes to collections of point processes. A preliminary conference version:

 

Applications

We have applied SES in various contexts, for example, to analyze steady-state visually evoked potentials and EEG responses to auditory stimuli, and to investigate the causal relation between morphological and molecular signaling events in cell migration; we are currently exploring additional applications. We refer to the following papers for more information on applications of SES:

·         J. Dauwels, T. Rutkowski, F. Vialatte, and A. Cichocki,
On the Synchrony of Empirical Mode Decompositions with Application to EEG,
Proc. IEEE Int. Conf. on Acoustics and Signal Processing (ICASSP), 2008.

 

Matlab Code

At present, Matlab code of SES for pairs of one- and multi-dimensional point processes is available. The current release is the version 1.0, which can be downloaded in the [Download] section.
About the latest updates, see the log file.

Matlab code for extracting "bump models", a kind of sparse time-frequency representations, can be found here; bump models can be used to describe electroencephalograms, magnetoencephalograms, and other kinds of time series. How to apply SES to bump models is described in:

In that paper, we use SES to detect loss of EEG synchrony in Mild Cognitive Impairment (MCI) patients.

 

If you have questions or comments, if the problem is not already discussed in the [FAQ] section, you can use the [Discussion group].
You can also directly [Contact us].

 

SES toolbox is licensed (creative commons license), and distributed for scholar use only.
SES toolbox 1.0 by Justin Dauwels et al. is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 2.1.
Publication of results obtained using this toolbox must cite relevant literature.