Manipulating individual neurons while maintaining their normal physiological functions is a crucial part of constructing a biological neural network with specific design synapse connections. Such networks are important for studying neurite outgrowth,

1
INTRODUCTION Manipulating individual neurons
while maintaining their normal physiological
functions is a crucial part of constructing a
biological neural network with specific design
synapse connections. Such networks are important
for studying neurite outgrowth, synapse
formation, and neural control in normal and
pathological conditions. In order to coerce
individual cells in these networks we have
developed a laser cell micropatterning system
that is capable of positioning individual cells
at submicron accuracy. In this paper we
concentrate on the development of a neurite
tracing algorithm that automatically assesses the
patterned cell culture based on that of Al-Kohafi
et al 1. OBJECTIVE Using our current
technique, individual cells are placed at
predetermined positions with submicron accuracy,
enabling detailed study of neurite outgrowths
regulated by various cell contact interactions
under phase contrast microscopes combined with
digital image processing techniques. In this
paper, the images of the cells are analyzed
automatically by an algorithm to detect the soma
and neurite outgrowth. Images of embryonic chick
forebrain neurons are used to demonstrate the
effectiveness of the technique. PRINCIPLE
PROCEDURE Optical forces generated by a weakly
focused laser beam are used to create a radial
trap of individual cells in the focal region of
the beam. The trapped cells are then guided
forward along the beam axis. Placement of a
movable substrate perpendicular to the beam axis
allows for the deposition of individual cells at
specific positions on the substrate. Movement of
the substrate, as the cells are trapped and
guided by the beam, allows for patterning cells
in a process known as laser guidance. After
pattern formation, the images are processed using
code written in Matlab. The soma detection is
based upon gray-level morphological processing,
while the neurite outgrowth tracing is based upon
the work of Al-Kohafi et al. 1, which is
modified to improve the computational
performance. Connected components are applied to
the morphologically closed binary image to yield
to the individual soma. High density of seed
points are initialized.
B
A
A two-step procedure is adopted to select
reliable seed points. Then the neurites are
recursively traced using normalized Gaussian
template. Each reliable seed point is traversed
in the forward and backward directions by
computing the score of the template at different
orientations. This is repeated until one of the
stopping criteria is met.
Laser Beam
Suspended Cells
Axial Force
Evenly Coated Substrate
Radial Force
Fig. 3 Soma and neurites detected by the algorithm
Pattern of Multiple Cell Types
Fig. 1 A Schematic depicting the principle of
laser guidance, B Patterned neurons with
intercellular spacing of 40 microns created using
the system. (20 microns
)
RESULTS Fig. 3 shows the traced neurites along
with the detected somas. Fig. 2 shows the image
of the neurites followed by the soma detected
image and the reliable seed points initialized
images. As shown in Fig 2.D and 2.E, the second
step removes a large number of false positives
that arise in the first step due to the gradual
intensity gradient in the upper-left corner of
the image. CONCLUSIONS We have described a
system to pattern neuronal cultures using a
weakly focused laser beam. The system is capable
of positioning individual cells at submicron
accuracy, thus opening the door for more precise
studying of neurite outgrowth, synapse formation,
and neural control in normal and pathological
conditions. We have also described an algorithm
for automatically analyzing the images of the
cells captured by our system. Future work should
be aimed at running the algorithm on a video
sequence to measure the outgrowth of the neurites
automatically. REFERENCES 1 K. Al-Kofahi et
al., "Rapid Automated three-dimensional tracing
of neurons from confocal image stacks," IEEE
Trans. on Information Technology in Biomedicine,
Vol. 6, pp. 171-187, 2002.
A
B
C
D

E
Fig. 2 A Image of neurons, B Closed image -
soma detected, C Seed points initialized, D
Reliable seed points detected using normalized
Gaussian kernels E Final reliable seed points
ACKNOWLEDGEMENTS South Carolina Spinal Cord
Injury Research Board SC BRIN Mr. Daniel
Bakken, Department of Bioengineering, Clemson
University.