Information Technology Reference
In-Depth Information
Like all imaging modalities, NSI is an endeavor that encompasses a great vari-
ety of multidisciplinary knowledge: modeling the neural electrophysiology of cell
assemblies, neuroanatomy, bioelectromagnetism, measurement technology, denois-
ing, reconstruction of time-resolved brain current flows from original sensor data,
and subsequent multidimensional analysis and interpretation in the spatial, tempo-
ral, spectral, and connectivity domains.
This chapter is an attempt to assemble these otherwise disparate elements in a
principled manner to provide the reader an in-depth overview of this exciting evolv-
ing field - with an emphasis on estimation techniques - that let us access functional
imaging at the speed of brain.
8.1.1 Neuronal Origins of Electromagnetic Signals
The measured EM signals that are generated by the brain are thought to be due
primarily to ionic current flow in the apical dendrites of cortical pyramidal neurons
and their associated return (a.k.a., volume) currents throughout the head tissues, i.e.,
the volume conductor [62, 63].
The unique architecture of each neural cell conditions the paths taken by both
the synaptically driven and intrinsic tiny intracellular currents that sum up vectori-
ally throughout the neuron to produce the dynamic net current generated by each
cell. This summation results in a significant net current at the cellular level if the
dendrites are organized along a single preferential direction rather than in a radial
shape. Furthermore, when multiple adjacent neurons with similar morphologies are
synchronously active, their cellular net currents constructively add up to produce
a group current density effect at the cell assembly level. For these reasons, assem-
blies of pyramidal cells in neocortical layers II/III and V are considered to be the
main sources of EM surface signals detected remotely. Neurons that have dendritic
arbors with closed field geometries (e.g., interneurons) are thought to produce no ex-
ternally measurable EM signals [37]. However, some non-pyramidal neurons such
as the Purkinje cells of the cerebellar cortex have been evidenced to generate EM
signals measurable at some distance [57].
Recent quantitative investigations using realistically shaped computer models of
neurons suggest that EM signals generated by neocortical columns made of as few as
50,000 pyramidal cells could be detectable outside the head and on the scalp. These
models also suggest that the contribution of intracellular currents due to voltage-
dependent ion channels involved in fast spiking activity might well be larger than
formerly expected, which supports the experimental evidence of high-frequency
brain oscillations (
100 Hz) detected from surface signals [54].
Although still somewhat controversial, there is cumulative evidence that activity
within deeper brain structures, such as the basal ganglia, amygdala, hippocampus,
brain stem, and thalamus [76, 96, 88, 42, 4], may be detected remotely. However,
single neurons produce weak fields, and if the current flow is spatiotemporally in-
coherent (e.g., a local desynchronization) the fields end up canceling. Thus, EM
>
