Molecular Bioimaging
Automated tracking and
analysis of moving objects in image sequences has been and continues to be one
of the major themes in digital image analysis research. This is not surprising
in view of its many applications in video surveillance, multimedia services,
automated vehicle guidance and
driver assistance, remote
sensing and meteorology, and medical imaging. It is also a recurring theme in
molecular biology. By their very nature, biomolecular systems are dynamic, and
it is one of the major chal-lenges of biomedical research and pharmaceutical
industries in the postgenomic era to unravel the spatial and temporal
relationships of these complex systems and to devise strategies to manipulate
them. Results in this area can be expected to have profound social and economic
impact in the near future, as they can be harnessed to improve human health and
well-being . Studies into biomolecular dynamics generate ever increasing
amounts of image data. To be able to handle these data and to fully exploit
them for describing bi-ological processes on a quantitative level and building
accurate mathematical models of dynamic structures, computerized motion
analysis is rapidly becoming a requisite.
Over the past decades, a
number of image analysis techniques have been developed in support of such
studies, the details of which were often buried in the small print of the
methods sections of papers published in the biology and biophysics literature.
The majority of these techniques were based on rather rudimentary principles,
however. The purpose of this article is to stimulate the application of more
advanced computer vision techniques to tracking in biological molecular
imaging, by surveying the literature and sketching the current state of affairs
in the field for a signal and image processing audience. After describing the
basic principles of visualizing molecular dynamics in living cells and giving
some examples of biological molecular dynamics studies, we summarize the
problems and limitations intrinsic to imaging at this scale. Then we discuss
the main ingredients of the commonly used tracking paradigm and subsequently
reconsider its competence by comparing it to certain aspects of visual motion
perception in human beings, keeping in mind the complexity and variability of
biological image data. We conclude by summarizing the main points of attention
for future research and the challenges that lie ahead.
No comments