Independent Component Analysis and Source Localization in

1

• Independent Component Analysis and Source
  Localization in
• Electrocorticography
• Kai Miller1, Rajesh Rao2, Donald Farrell3, and
  Jeff Ojemann4
• University of Washington,
• Dept. of Physics 2) Dept. of Computer Science 3)
  Dept of Neurology 4) Dept.of Neurosurgery
• Contact Kai Miller kjmiller_at_u.washington.edu

• Introduction and Background
• There are many ways to assess neurological
  function in the brain, using methods ranging from
  behavior, to cell and tissue pathology, to MRI
  and other forms of imaging, and to direct
  measurement of the neural signal propagation in
  the form of varying potentials.
• On the macroscale and without compromising the
  integrity of the brain, a bulk potential can be
  measured at various points outside of the brain,
  and is the result of the combination of many,
  many single neurons exhibiting coherent behavior.
  These phenomena have been studied extensively
  from the surface of the head with
  electroencephalography (EEG), but the underlying
  fields and currents giving rise to these
  potentials are smeared heavily, both spatially
  and temporally, as they pass through the
  meninges, calvaria, and scalp because of the
  capacitative, resistive, and inductive properties
  of these tissues.
• For this reason, the ability to study these
  signals subdurally, from the surface of the
  brain, is of particular appeal. With
  electrocorticocraphy, we have the rare privilege
  of being able to study a patient population with
  arrays of electrodes placed directly on the brain
  surface (Electrocorticographic ECoG implants)
  in the week prior to a cortical resection for
  removal of epileptic foci.
• Independent component analysis is a way of
  isolating events on the surface of the brain
  which are statistically independent of each
  other. We plan to use these components to
  examine bulk electrical properties from the
  surface of the brain for the purpose of
  understanding the brain, and for the purpose of
  creating a brain computer interface (BCI) which
  subjects can consciously and quantitatively
  modify.
• For paralyzed patients, this has the potential
  for surrogate control of muscle via a
  brain-computer-muscle connection. This has been
  done, using EEG, but because of noise from
  cranial muscle and inherent noise in the EEG,
  success has been limited. BCI for the purpose of
  communication, using EEG, has been tried by
  several groups, with visual and auditory
  feedback, and is promising, but the higher
  fidelity signal from ECoG may permit more degrees
  of freedom and a signal less susceptible to
  noise.

References   Eric C Leuthardt, Gerwin Schalk,
Jonathan R Wolpaw, Jeffrey G Ojemann and Daniel W
Moran, A braincomputer interface using
electrocorticographic signals in humans, Journal
of Neural Engineering, vol. 1, no 2 pp. 63-71,
June 2004   A. Delorme and S Makeig EEGLAB an
open source toolbox for analysis of single-trial
EEG dynamics J Neurosci Methods, 1349-21,
2004.   Pierre Comon, Independent component
analysis A new concept? Signal Processing,
vol.36, no.3, pp.287-314, April 1994.   A. Bell
and T. Sejnowski An information maximisation
approach to blind separation and blind
deconvolution. Neural Computation, 7, 6,
1129-1159, 1995
Seizure spread in the array is shown. ICA picks
out events like the seizure focus.