BioS: a New Tool for Biopotential Experiments

Share Embed


Descrição do Produto

Proceedings of the 29th Annual International Conference of the IEEE EMBS Cité Internationale, Lyon, France August 23-26, 2007.

SaC02.1

BioS: a New Tool for Biopotential Experiments Maide Bucolo Member IEEE, Federica Di Grazia Student Member IEEE, Mattia Frasca Member IEEE, Luigi Fortuna Member IEEE, Francesca Sapuppo Student Member IEEE. Abstract— BioS (Bio-potential Study), a novel environment for Biopotential analysis is here proposed. It provides the processing of biopotentials of any number of channels distributed with different geometry. It includes several features as data importing, data visualization (1D, 2D, 3D), preprocessing (frequency & saturation filtering, statistical analysis), spatio-temporal processing (power spectrum analysis, nonlinear analysis, independent component analysis both in the spatial and time domain). BioS, also, provides a user- friendly Graphic User Interface designed to allow all user to speed up data analysis experiments. I. INTRODUCTION he development of innovative analysis techniques in bio-engineering has had a great impact in the medical field. The realization of new instruments and the use of new methodologies turn out to be of a great advantage for the study of biopotentials in which the variability and the uncertainty affect both the space and the temporal domain. The resulting increase of the available data has had a great influence on the implementation of modular and portable software environments able to an easy data analysis. In literature many tools for biological signals analysis are already described and used for different applications. Most of them are developed under the Matlab environment. Examples of commercial solutions [1] are PRANA, EMSE®, while some free-ware products [2] are BrainStorm and EEGLAB. The software BioS (Bio-potential Study), here described, was also realized in Matlab®7 as integrated development environment (IDE) for its flexibility, and the possibility of programming by scripts using pre-existing toolboxes. The effort in the development of BioS was to create a modular and portable environment for the acquisition, the visualization, the preprocessing and the processing of any biopotential data as described in section II. It was thought to give the possibility to analyze various kind of data (EEG, MEG, ECG, EOG) with the parametrization of different instrumentation characteristics. The aim is to create an environment that easily gives the possibility to integrate or to explore new linear and nonlinear methodology, as well as the possibility of testing the already developed methods on different cases of study. BioS supports a graphic user interface (GUI), shown in

Fig.1, designed to allow non-experienced Matlab users to apply advanced signal processing techniques to their data.

T

Manuscriptum received April 2, 2007 M. Bucolo, F. Di Grazia, M. Frasca, F. Fortuna, F. Sapuppo are with the Dipartimento di Ingegneria Elettrica, Elettronica e dei Sistemi, Università degli Studi di Catania, V.le A, Doria 6, 95125 Catania, Italy. (e-mail: mbucolo.diees.unict.it).

1-4244-0788-5/07/$20.00 ©2007 IEEE

Fig.1 BioS main Graphic User Interface and its actual potentialities.

II. METHODS AND RESULTS A. Data Structures and Acquisition Due to the wide range of instrumentations available in the market, the main issue dealing with the implementation of the acquisition system is the parameterization of several characteristics of both the acquisition protocols and the biopotential signals. A single data structure (‘Bios_info’) has been created to store parameters related to the acquisition system (sampling rate, overhead) or the data (number of acquisition channels, number of minutes, total recorded time). Moreover it is possible to acquire signals in different data format as binary (Little-endian or Big-endian), ASCII and Matlab. The Bio_info structure can be also accessed directly from the Matlab command line to extract processing history and data info. The analysis results are given in a Matlab type and can be saved in different format as required by the user. Different data bring different spatial distributions, therefore the system provides the possibility of inputting the coordinates of the acquisition points organized in files (with an opportune protocol) to reconstruct the geometry of the system. The environment has been tested on Magnetoencephalography (MEG) data acquired with two different systems: whole-head 148-channels (4-D Neuroimaging, San Diego, California) MEG instrument and dual 37-

5190

channel bio-magnetometers (BTI, Inc.) both located at The Scripps Research Institute (La Jolla,CA). A 3D representation of the MEG head channels distribution is shown in Fig.2.

a

b

Fig.2 Head channel distribution for: a. whole- head 148 channels b. dual 37channel bio-magnetometers.

The spatio-temporal maps is another kind of output representation supported by BioS tool. Fig. 3d illustrates, as example, the evolving averaged results over one minute intervals of entire whole-head MEG. The color of the pixel (jth ,ith) represents the value of the resulting analysis related to the jth minute for the ith channel, using a conversion that is described by the color-bar on the right of the image. Thus, the image's ith column represents the time evolution by one-minute time-windows at the ith channel. The advantage of this representation is the possibility to have a complete view of the evolving in time of the parameter under study for all channels, loosing not the spatial information but just the spatial distribution. C. Preprocessing

B. Visualization. Using the pre-loaded files during the acquisition feature it is possible to have different visualization modalities: onedimensional, bi-dimensional and three-dimensional. This represents a great potential available not only for the raw data but also for the pre-processed and processed data.

a

b

It is well known that the data can be corrupted during the acquisition due to several reasons (broken channel/s, background noise). According to this problem, preprocessing methods are included providing standard data analysis functions organized in two different sections: one is the frequency and saturation filtering and the other is the statistical analysis. Regarding to the filtering, in order to develop a faithful support to the user, it is possible to project different filters on few channels data and to extend the filtering to all data, simultaneously. Fig.4a shows a zoom of the raw time series dynamic in the grey line and the filtered signal in the black line.

c d Fig. 3 a. One Dimensional visualization b. Two dimensional visualization c. Three dimensional visualization.d Spatiotemporal map.

In Fig.3 , an example of 1D, 2D and 3D plots are reported. In particular the spatial maps (2D and 3D) are mapped using a color code for any output signal at the ith intersection point of the mesh on the digital reconstruction of the coils positions, where i = 1, 2, 3,…, n (n=number of channels). The color values in the intermediate areas of the mesh are obtained through linear interpolation between neighbouring points. As shown in Fig.3.b, it is possible to scroll 60 head maps related to each second of a single minute by using the bar at the bottom of the canvas. The information has been averaged choosing the second as range of interest, but it would be possible a parametrization of the study range in order to push the study to the millisecond range, that is the time interval of the neuronal activity.

a

b Fig. 4 a. Frequency filtering of a raw data b. Saturation filtering of a raw data.

This procedure can be applied both with a frequency filtering and a saturation filtering. The saturation filtering is particularly suitable feature for the detection of broken channels. The choice of the parameters for the saturation

5191

filtering implementation permits to decide any value of the amplitude threshold in which delimitate the signal. As shown in Fig.4b the red lines represent the saturation threshold over and under which the signal has been assigned to a desired value. The statistical analysis section has been thought to apply multi-statistical methods looking at “inside signal”, “between signals” and “among signals”. The inside signal panel gives the statistical information about a selected channel (mean, standard deviation, histogram) and plots its state space representation, meanwhile between signals extracts information related to the cross-correlation signal between two channels (maximum value and respective delay, value at zero lag) and plots their state space representation. The among signal panel permits to have a spatial map of the statistical parameters over a selected time range; to plot the histogram related to the mean and standard deviation of the channels; and finally the spatial maps for the three parameters extracted from the cross-correlation between one channel with all the others. In Fig.5 some examples of statistical analysis are reported: (a)one channel signal histogram, (b)crosscorrelation between two signals, (c)state space representation of two signals, and (d)spatial map visualization of cross-correlation maximum.

a

b

For brevity only some explicative results have been chosen and reported in order to support the potentiality of Bios. The power evaluated over one-minute period is reported in Fig. 6 in the whole-head map(a) and dual-head map(b) visualization, meanwhile the spatio-temporal maps of both geometry are shown in Fig. 7.

a b Fig.6 Spatial maps. a. whole- head power map. b. double power map.

a b Fig.7 Spatio-Temporal map: a. whole- head geometry. b. dual area geometry.

Regarding to the ICA analysis an exiting Matlab tool, FastICA [5], has been used and integrated to the BioS environment. The ICA attempting to separate data into maximally independent groups in time or space, yielding temporal-ICA (TICA) and spatial-ICA (SICA) respectively. In Fig.8 the BioS output for TICA analysis over one minute is reported and it shows a clear presence of the heart beat signal.

c d Fig. 5 a. Signal histogram b. Cross-correlation analysis between two signals c. Signal State Space representation d. Space maps.

D. Processing The nonlinear nature and the space extension of the biological systems under investigation makes physiological processes analysis similar to the one associated to the extended complex systems theory. Three analysis methods were integrated in the processing module: the first one is relative to the spatio-temporal power distribution, the second to the spatial and temporal independent components analysis [3] and the third concerns the nonlinear analysis methods for the extraction of the maximum Lyapunov exponent (λ) and d-infinite (d∞) parameter [4] in the spatio-temporal domain. The processing actions can be applied both on raw data and preprocessed signals.

Fig. 8 Temporal independent component analysis.

The SICA method has been used to extract spatial information as Spatial Modes (SMs)[6] and to group them by performing a suitable clustering strategy that means to identify a set of basic spatial patterns whose linear combination represents the spatial patterning of the whole system. Fig.9 shows how the clustering approach permits to extract Similarity Modes(a) and Dissimilarity Modes(b) in a comparison of different minutes.

5192

a

b Fig. 9 a.Similarity modes b.Dissimilarity modes in Biopotentials patterns.

Nonlinear methods have been implemented for the extraction of indicators such as d∞ and λ using a computational algorithm [7] resulting less onerous than the conventional ones [8]. The output representation can be obtained, even in this case, in the spatio-temporal and spatial maps for any geometry as show in Fig.10 a and b respectively..

both useful for next studies. The already integrated processing methods are the power spectrum analysis, the spatio-temporal non linear analysis and the independent component analysis in the spatial and temporal domain. Both preprocessing and processing results can be visualized in spatial images. BioS features have here provided to be effective by testing on MEG data with two different instrumentation and geometry and has been tested the possibility to work with different type of data (EEG data in txt format).It’s already planned the possibility to add a section dedicate to the multiple data set analysis (such simultaneously recordings of nasal cycle, blood pressure, EEG). The analyses provided by Bios on the biosignals can give a support to a successive medical interpretation of the data for clinical purpose. Moreover future development could be the porting of Bios code on grid computing architecture, parallelizing the processing of data in the space (areas or channels) and time (minutes or seconds or mseconds) domain characterizing the nature of biopotential data in order to provide a statistical analysis giving a clinical support creating a Data Base for diagnostics analyses. ACKNOWLEDGMENT Thanks to D. Shannahoff-Khalsa for his fruitful collaboration working on “Analysis of MEG data for treatment of Obsessive Compulsive Disorder with yoga breathing technique”. MEG data provided by the Institute for Nonlinear Science, University of California, San Diego, La Jolla, California, USA.

REFERENCE a b Fig.10 a. Spatio-temporal d∞ maps. b. Average d∞ head maps.

III. CONCLUSIONS The BioS software developed in Matlab®7 is a userfriendly interactive environment allowing biopotential data processing and results representation through images. The software architecture has been designed to be modular and flexible in order to meet any future requirement becoming a suitable instrument for a medical reading of the results. This software allows to processes biological data from different acquisition systems, characterized by different sensor distribution geometry (whole head, multiple areas), by different data format (binary, ASCII and Matlab). The analysis provided by BioS could help both non experienced users to study biopotential data without programming and experienced users to speed up the analysis of the experimental data. The BioS features are data visualization, preprocessing and processing. Regarding to the visualization BioS allows a onedimensional, two-dimensional, three-dimensional and spatio-temporal visualization not only for the raw data, but also for the treated signals. The BioS preprocessing feature permits to perform a frequency & saturation filtering and a statistical analyses

[1]http://www.phitools.com/index.html, http://www.cortechsolutions.com/ [2]http://www.sprweb.org/repository/index.html, http://neuroimage.usc.edu/brainstorm/ [3] A. Hyvarinen and E. Oja, “Independent component analysis: algorithms and applications”, Neural Networks, 2000, no. 13, pp.411-430. [4] F. Sapuppo, E. Umana, M. Frasca, M. La Rosa, D. Shannahoff-Khalsa, L. Fortuna and M. Bucolo, ‘Complex SpatioTemporal Feature in MEG Data’, Mathematical Biosciences and Engineering, October 2006, Vol. 3, No. 4, pp. 697-716. [5] The FastICA package is Copyright (C) 1996-2005 by Hugo Gävert, Jarmo Hurri, Jaakko Särelä, and Aapo Hyvärinen. [6] G. Bucolo, M. Bucolo, M. Frasca, M. La Rosa, D. ShannahoffKhalsa, M. Sorbello, “Spatial Modes in Magnetoencephalography Spatio-Temporal Patterns.”Proceedings of the 25th Annual International Conference of the IEEE EMBS, Cancun, Mexico, 2003. [7] S. H.Strogatz, “Nonlinear dynamics and chaos: With applications to physics, biology, chemistry, and engineering.” Reading, MA: Perseus Books, Cambridge MA. 1994. [8] F. Sapuppo, E. Umana, M. Frasca, M. La Rosa, D. Shannahoff-Khalsa, L. Fortuna and M. Bucolo, ‘Complex SpatioTemporal Feature in MEG Data’, Mathematical Biosciences and Engineering, October 2006, Vol. 3, No. 4, pp. 697-716.

5193

Lihat lebih banyak...

Comentários

Copyright © 2017 DADOSPDF Inc.