Digital techniques to extract information from EEG patterns.

ebernard at ebernard at
Wed Mar 6 17:18:21 EST 1996

 Hello neuronetters,
  Continuing an earlier discussion in this group I would like to pay attention  to modern
  mathematical techniques to analyse EEG patterns, which have become available 
  since digital  techniques became feasible with workstations, fast personal computers, large memories and
  cheap digital storage media.
  I hope you neuronetters excuse me to have taken some time to get my ideas on paper as good
  as I can.

  - First I will give my ideas about the new analysis methods of EEG data and personality,
  - Then I will give information on new analysis methods for the EEG
  - I will support this with information on the Neurometrics Method
  - I will also pay attention to research based on Chaos Theory and EEG,
  - I will give a short list of references to these methods.
  - I will conclude with abstracts of the preceding discussion.

  I agree with the principal viewpoint of David Seaman, that "person" is a concept, that has
  been created by humans. It has no relevance to the real world other than how it is
  defined. I also agree with Mariela Szirko, that the attempt to apprehend the traits of
  personhood as used by our society to determine the rights of entities is a sociological
  study. It can be dangerous to try to connect "personality" to EEG brainwave patterns.
  Besides that lies the question if it is possible to identify on the basis of EEG data
  whether individuals belong to a certain group.

  A problem to be solved first is how to define membership of such a group or species in
  such a way that it has any relevance to the notion of the group in the real world. Suppose
  that is possible (and it has been showed to be possible in various cases).

  From what I have read about Neurometrics and Chaos Theory methods I think it will be
  possible with computer analysis of the digital EEG to find differences between groups of
  individuals or species, where the group definition is meaningful in relation to its notion
  in the real world. With neurometrics, significant differences in brain activity have been
  reported for instance between painters after a 20 year career and young painters and
  between 'normal' man and woman of the same age. The effects of technical problems, like
  noise, biasing and aliasing can be excluded effectively.
  I will support this idea with some information and references on these methods in the
  following article.
  I do not know if studies are done to establish with reasonable confidence the sex, age,
  species or other non-EEG feature of one concrete EEG (interesting research suggestion?).
  What I read is that sex and age are important variables when EEG data are analysed
  with the Neurometrics method.
  What I also read is that it is possible with Neurometrics Method to determine significant
  differences in brain function between a diseased individual and the EEGs of a group
   normals' with same age and sex.  The examination with Neurometrics is done on one EEG of
  the individiual who is diseased in some way, while the differences can not or hardly be
  seen with the bare eye on the EEG.

  But still, at this moment, the question what the relevance of the observed differences
  might have to whatever concept of personality remains unanswered.

  With the digital EEG combined with various kinds of mathematical techniques on computers,
  EEG analysis enters a new era. Costs are low and the scientific possibilities are
  challenging. The underlying techniques are proven and reliable.

  For some years now it is feasible to make digital multichannel recordings of the EEG
  patterns with workstation-like computer configurations, since large processor capacity,
  memory and storage capacity became affordable.

  Digital recording of the EEG ("Paperless EEG") has the advantage that storage of the
  digital recordings is much cheaper and less room-consuming than paper EEG's. This cost
  benefit only already rectified the purchase of a digital EEG system for most institutions
  in my country.
  Besides the cost argument, the digital EEG offers very interesting possibilities from
  scientific viewpoint. In the conventional paper recording the technical help to the
  neuroscientist was limited to amplification and filtering. Once chosen and recorded, it
  could not be changed. Nowadays, with digital EEG, it is feasible to use the enormous
  reservoir of mathematical techniques, some existing for decades already, to analyse the
  EEG data as often as one whishes. Almost any filter one can think of can be applied.
  Since decades also the possible relevance of mathematical analysis to EEG data has being
  researched. A vast amount of litterature on the subject has been published and very
  interesting and useful results are reported.
  Various fields of mathematics contribute with useful methods for EGG analysis like
  statistics, (e.g. multivariate analyses, neurometrics, see JOHN83, JOHN90, JONK93, WEER93),
  pattern recognition, topography
  (brain mapping), chaos theory (see STAM94, STAM95, JELL95), integral transformations (like
  Fourier). The dramatically increased power and decreased prices of workstation-like
  computer configurations and commercially available software with user  friendly Windows
  interfaces, nowadays bring the advantages of digital EEG analysis under the
  reach of health institutions outside the "rich" research world.

  With the digital EEG combined with various kinds of mathematical techniques on computers,
  EEG analysis enters a new era. From a mandatory routine exercise the EEG can become a
  modern instrument with challenging possibilities and fascinating results.
  For the neuroscientific relevance of every approach you must refer to the appropriate
  litterature, since I am not a neuroscientist. In this article I will limit to two examples
  of new possibilities with the digital EEG and some references.

  Based on my expertise in information theory, electronic engineering and informatics
  engineering I can assure you that it is possible to process properly recorded and
  digitized EEG data in such a way that side effects (noise, aliasing, biasing etc.) are
  controlled and render reliable results.

  Of course, the means used in recording and processing the digital EEG, like electrodes,
  cables, amplifiers, samplers, digitizers, processors, storage devices, analysis algorithms
  and their implementation, must be of controlled quality, complying strict specifications.
  Therefor it is important to rely on proven configurations and software (like e.g.
  "Brainstar" from Schwind Benelux Medical Electronics, Oosterbeek, The Netherlands,
  Tel +31 26 3333744).

  The Neurometrics method is based on the method of dr. E.R. John for analysis of the EEG.
Certain data, the neurometrics parameters, of an individual are compared to those of a reference
file of individuals within the same age range.

The Neurometrics method is meant to recognize details in the EEG that can not, or only by very few
highly specialized people, be seen in the EEG with the bare eye. For severe cases of brain disease
the Neurometrics method is not necessary or useful. A well known application of Neurometrics is
diagnoses of change of a patient in time, for example before and after an operation.

The Neurometrics methode works as follows.

More information about the Neur-sci mailing list