Human Brain Mapping
Florence, Italy
June 11-15, 2006
Rey R. Ramirez & Scott Makeig , Institute for Neural Computation, University of California San Diego, La Jolla CA
Independent co-modulation of spectral EEG power in subsets of independent brain sources
INTRODUCTION: A new framework for neuroelectromagnetic inverse imaging is developed in which the fundamental atoms of data representation are magnetic field vectors generated by locally distributed multi-resolution current density geodesic functions centered at each source point, instead of the traditional gain vectors generated by independent dipoles. A highly overcomplete lead-field dictionary of magnetic field patterns is constructed using these distributed neural bases at multiple scales. Cortical geodesic distances and orientation constraints are exploited to discriminate between gyri or sulci that are near each other in Euclidian space but not in geodesic cortical space. A sparse solution to the multiscale-transformed overcomplete subset selection problem is computed using a generalized sparse Bayesian learning (SBL) algorithm. Although this solution is maximally sparse in terms of the transformed system of equations, it represents a distributed current density estimate once it is back-transformed with the a priori neural basis matrix. Results show that this algorithm can perfectly reconstruct many simultaneously active sources with different spatial extents. It can also separately localize each sensor map learned, from the unaveraged data, by any Independent Component Analysis or Blind Source Separation algorithm.