Gao, Yongxiang; Kilfoil, Maria L.
Description
Localization and tracking of colloidal particles in microscopy
images generates the raw data necessary to understand both the dynamics
and the mechanical properties of colloidal model systems. Yet, despite
the obvious importance of analyzing particle movement in three dimensions
(3D), accurate sub-pixel localization of the particles in 3D has received little
attention so far. Tracking has been limited by the choice of whether to
track all particles in a low-density system, or whether to...[Show more] neglect the most
mobile fraction of particles in a dense system. Moreover, assertions are frequently
made on the accuracies of methods for locating particles in colloid
physics and in biology, and the field of particle locating and tracking can be
well-served by quantitative comparison of relative performances.
We show that by iterating sub-pixel localization in three dimensions, the
centers of particles can be more accurately located in three-dimensions (3D)
than with all previous methods by at least half an order of magnitude. In addition,
we show that implementing a multi-pass deflation approach, greater
fidelity can be achieved in reconstruction of trajectories, once particle positions
are known. In general, all future work must defend the accuracy of
the particle tracks to be considered reliable. Specifically, other researchers
must use the methods presented here (or an alternative whose accuracy can
be substantianted) in order for the entire investigation to be considered legitimate,
if the basis of the physical argument (in colloids, biology, or any
other application) depends on quantitative accuracy of particle positions.
We compare our algorithms to other recent and related advances in location/tracking
in colloids and in biology, and discuss the relative strengths
and weaknesses of all the algorithms in various situations. We carry
out performance tests directly comparing the accuracy of our and other
3D methods with simulated data for both location and tracking, and in
providing relative performance data, we assess just how accurately software
can locate particles. We discuss how our methods, now applied to colloids,
could improve the location and tracking of features such as quantum dots in
cells.
Items in Open Research are protected by copyright, with all rights reserved, unless otherwise indicated.