Neuronal 3D distribution reconstruction
system based on a priori morphological model and series of real 2D-images of
brain sections
Yu.
V. Obukhov, V. E. Antsiperov, O. V. Evseev
Kotel'nikov Institute of Radio-engineering and Electronics of RAS
Received
July 23, 2012
Abstract. The paper presents results
of the development of reconstruction algorithms of three-dimensional distributions
of neurons, previously recognized by microscopic images of brain slices of
experimental animals. The adaptation of the well-known EM-clustering algorithm to
highlight groups of neurons in slices is considered, taking into account the
prevailing today neurobiological concepts. Authors propose a polynomial
interlayer approximation of the parameters of two-dimensional contours of
groups. It provides reduce at least halve the number of slices for the
three-dimensional reconstruction of neuronal distributions. Described
algorithms are implemented in a software package developed by the authors for constructing
neuronal 3D-distributions on a series of 2D-slice images of the brain. It is
supposed to be a useful tool for research in present neuroscience, in
particular, for problems of the modeling of Parkinson's disease, and will
significantly reduce the costs of such researches.
Keywords: 3D distributions reconstruction,
2D clustering, EM-algorithm, B-splines.