Event Abstract

Back to Event

Iterative N-way PLS for real-time control of external effectors with ECoG recordings

Andrey Eliseyev1, 2, 3*, Cecile Moro2, Napoleon Torres2, Thomas Costecalde2, Sadok Gharbi4, Corinne Mestais2, Alim Louis2, 3 and Tatiana Aksenova1, 2
  • 1 Foundation Nanosciences, na, France
  • 2 Clinatec, LETI/CEA, France
  • 3 Joseph Fourier University of Grenoble, EDISCE, France
  • 4 LE2S, LETI/CEA, France

Brain Computer Interface (BCI) aims to provide a way for effectors control based on measurements of brain electrical activity. Majority of current BCI systems is driven by the external cues to distinguish informative data periods from brain’s general functioning (cue-paced BCI). This restriction seems to be extremely burdensome in the real-life applications. As opposed to cue-paced BCIs, self-paced ones do not use any cues. However, their current implementations suffer from high level of false activations. Thus during experiments subjects have to be concentrated on the task which limits real-life application of self-paced BCI systems.

In general, BCI experiments consist in two stages. The first one, called calibration, is to identify dependences (for instance, linear regression) between the brain data observations and external device state indicator y. The second stage consists in application of the constructed model for prediction of y(t) and corresponding effector activation during the continuous monitoring of neuronal activity.

Multimodal analysis gives opportunity to analyze simultaneously temporal, frequency, spatial, etc. dynamics of the registered signal. Thus it presents a promising system calibration approach to improve quality of self-paced BCIs. Multimodal data are effectively represented by means of tensors (multi-way arrays), which are higher-order generalization of vectors and matrix concepts. N-way Partial Least Square (NPLS) algorithm is proposed for constructing a linear model of the depending variable (effector state indicator) y(t) and independent variable (tensor of Fourier or Wavelet transforms of brain neural activity recordings) x(t) at the moment t. The method is particularly suited for those observation tensors X which contain highly correlated variables and/or amount of observations considerably less than variable’s dimension. Moreover, NPLS saves tensor data structure that improves robustness of results. A drawback of the method is its huge storage consumption for holding of observation tensor X in the active memory. To overcome this problem we proposed an iterative NPLS algorithm (patent pending). It is based on fragmentation of the initial dataset on subsets and their sequential treatment. Thus at every moment only a small part of the data is stored in the active memory. It allows treating datasets of huge dimension. Performance comparison of INPLS vs. generic NPLS was carried out on the artificial data set. Obtained results show that iterative algorithm preserves the accuracy of NPLS. Moreover it demonstrates better robustness and overfitting suppression.

The developed method was tested with a set of self-paced BCI experiments in the freely moving rats over more than 6 month. Every experimental session was lasting up to an hour. Electrocorticography (ECoG) signals are recorded by a set of electrodes implanted on the cortex surface. Then signal was filtered and mapped to the temporal-frequency-spatial space with continuous wavelet transform, to form the observation tensor X. Two-position pedal was used as an effector. The results are characterized by high level of detection rate, whereas false activation rate was remaining acceptably low. Thus INPLS algorithm represents a promising tool for self-paced BCI systems in real-live environment.

Keywords: computational neuroscience

Conference: Bernstein Conference on Computational Neuroscience, Berlin, Germany, 27 Sep - 1 Oct, 2010.

Presentation Type: Poster Abstract

Topic: Bernstein Conference on Computational Neuroscience

Citation: Eliseyev A, Moro C, Torres N, Costecalde T, Gharbi S, Mestais C, Louis A and Aksenova T (2010). Iterative N-way PLS for real-time control of external effectors with ECoG recordings. Front. Comput. Neurosci. Conference Abstract: Bernstein Conference on Computational Neuroscience. doi: 10.3389/conf.fncom.2010.51.00127

Copyright: The abstracts in this collection have not been subject to any Frontiers peer review or checks, and are not endorsed by Frontiers. They are made available through the Frontiers publishing platform as a service to conference organizers and presenters.

The copyright in the individual abstracts is owned by the author of each abstract or his/her employer unless otherwise stated.

Each abstract, as well as the collection of abstracts, are published under a Creative Commons CC-BY 4.0 (attribution) licence (https://creativecommons.org/licenses/by/4.0/) and may thus be reproduced, translated, adapted and be the subject of derivative works provided the authors and Frontiers are attributed.

For Frontiers’ terms and conditions please see https://www.frontiersin.org/legal/terms-and-conditions.

Received: 23 Sep 2010; Published Online: 23 Sep 2010.

* Correspondence: Dr. Andrey Eliseyev, Foundation Nanosciences, na, Grenoble, France, andreyel@gmail.com