Health Telematics (AIM) Final Report


	Updated: Feb 1, 97
ESPIS

EEG signal processing and interpretation during sleep

Project code: A2008

Project value: 898 KECU

EC contribution: 450 KECU

No of partners: 4

No of countries: 3

Duration: 24 months

Contact:
Mr. Michel Toussaint
FORENAP
Centre Hosp. Spec. En Neurosciences
Pavillon 4
F-68250 Rouffach, France
Tel.: +33-89-49.60.51 ext. 16.80
Fax: +33-89-78.51.26

Project overview

The project's aim was to develop a dedicated workstation in order to process multiple channels of electrophysiological signals in real-time during sleep. In ESPIS we were aiming to define both an architecture and an environment for EEG signal interpretation in medicine based on computer science gold standards (UNIX, XWindow, Motif). The design takes advantage of the development environment, processing and graphics resources, and networking capabilities of high performance UNIX workstations. Signal processing and pattern recognition analysis are provided by parallel processing on a specific developed acquisition architecture (DSP) based on transputers.

Purpose and objectives

The proposed architecture is designed around a classical workstation with high graphic capabilities, able to manage and store data and to communicate through networks. On the other hand a high performance DSP board based on a transputer is plugged in the workstation.

We specified a numerical toolbox running efficiently in real-time on the DSP board with the following objectives: definition of the digital signal processing subroutines for signals processing during sleep, definition of statistical data analysis procedures in order to characterise the main signals' properties (feature extraction) and definition of pattern recognition methods to classify EEG and sleep patterns. Data reduction coming from digital signal processing techniques to a small meaningful subset of relevant "features" is the second step in the analysis of sleep signals, before taking the decision. finally, pattern recognition algorithms are used to compute a mathematical rule which classifies a data set into a clinical or an experiment category.

Results

The main result is a high performance prototype demonstrating signal interpretation during sleep which has already been tested in a medical environment (3 clinical centres and 1 research centre). The overall specifications allow this biomedical device to be extended to other types of medical signals.

The proposed system has a modular architecture taking into account the opening and portability through the use of standards. It is built around the main following elements: a UNIX workstation (68040 - 32 MBytes), and EEG signals acquisition and processing sub-system (32 channels DSP board), a data mass storage unit (1.3 GBytes hard disk, 5 GBytes DAT) and high resolution graphic units (19" Colour XWindow terminals) connected to the workstation through Ethernet. Each sub-system can be adapted according to the needs of the ultimate user.

The numerical toolbox processing routines are available through menu-driven graphic interface and can be applied in real-time on each channel independently.

The exploitation of results has been along the following lines:

Re-assessment of the market potential and of the markets to be addressed in the exploitation phase

Thanks to a lot of extra efforts made to industrialise the prototype, we are now able to show a first product which can be used not only by researchers and computer scientists but also by medical personnel. This has allowed TELMAT to respond to a certain number of calls for tender which where given out in the last six months by hospitals in France.

Despite the fact that no smart marketing presentation of the product exists nowadays and taking into account that world leaders (Nicolet, Micromed, Nihon Kohden ...) were present, our proposal was in all cases considered as one of the top two or three proposals. It appears that the foreseeable advantages of our solution are confirmed and correspond well to physicians present needs. Our concept is particularly well adapted to relatively important EEG departments in which a tendency is to install a modern computer structure in which individual (micro)computers will be connected to a server usually containing a large data base.

TELMAT's experience in supplying such client/server solutions is an important advantage in comparison to competitors.

However, each company which sells computerised EEG systems also propose systems for small clinics. Unfortunately, for these clients the openness of the system is secondary in comparison to economic criteria. This brings us to study a low cost version of the application able to run on PC hardware.

Commitment and ability of the participants to assure the transfer of the results into practical and effective use

For six months many of the calls for tender published in France were received in the domain of computerised EEG. Thus we have established that we are able to construct a good dialogue with the laboratories which want to use the system for research and who have a certain experience in computer science.

On the other hand it seems that it is more difficult for us to sell directly in the medical environment where our experience is insufficient. For this reason we have decided to find a biomedical company which is interested in distributing the product.

For the moment, two French companies RACIA-ALVAR and TGS, well known in Europe in the EEG field and the medical field, are very interested in a collaboration and some preliminary meetings have already been organised with them.

efforts so far undertaken and to be undertaken in the future to assure transition to a successful exploitation phase

It is well known that there is an important gap between a research prototype ad a commercial product. A lot of effort must be put on the user interface and on all of the marketing aspects.

During the last six months, all calls for tender published in France by public health care organisations, allow us to know precisely what the market wants and in fact what is missing in our product to be the most competitive. A a first set of commercial documentation has been prepared.

Future activities:

to finalise the industrialisation of the present product,
to develop complementary functions (video EEG, evoked potentials, ...)
to increase the hardware product range (portable system, PC version ...)

List of Deliverables

Year 1

State of the art report (3/R/L)
Requirements for low and intermediate levels report (9/R/I)
Pre-prototype (Acquisition and monitoring of signals) (9/P/I)

Year 2

Signal processing toolbox (6/T/R)
Criteria specification report (3/R/R)
Learning set (6/O/I)
Prototype for testing (6/P/I)
Performance and validation report (12/R/P)

List of Participants

Mr. Christian Pflieger Telmat Informatique 6, rue de l'Industrie F-68360 Soultz, France Tel.: +33-89-76.51.10 Fax: +33-89-74.27.34	Dr. Axel Steiger Max Planck Inst. für Psychiatrie Klinisches Institut Kraepelinstrasse 10 D-8000 München 40, Germany Tel.: +49-89-30.62.22.36 Fax: +49-89-30.62.20.00
Dr. Teresa Paiva Centro De Estudos Egas Moniz INIC Lab. EEG Hopital de Santa Maria P-1699 Lisboa, Portugal Tel.: +351-1-769.50 Fax: +351-1-76.40.59



	Copyright 1997 © EHTO All rights reserved This server is the only official EHTO WWW knowledge repository. Mail suggestions to: webmaster@ehto.org