Kinetochore microtubules and the associated
plus-ends have been under intensive investigation in cell biology and
molecular medicine. Though electron tomography provides new possibilities
in imaging their high-resolution structures, the interpretation of the
acquired data remains difficult because of the complex and cluttered
cellular images. As a result, practical segmentation of the tomography
volume has been dominated by manual operation, which is time consuming and
subjective. In this research, we develop a model-based automated approach
to extracting tubular objects and curvilinear structures from noisy volume
images. In particular, we make the segmentation robust by systematically
incorporating various prior models of the structures throughout image enhancement
and segmentation. Our segmentation methods, regularized under the prior
constraints, successfully extract fine biological structures such as
microtubules and their plus-ends from highly complicated and noisy cellular
environment.
Experimental
results (click for sample video) demonstrate
that our automated method produces results of microtubules and plus-ends
comparable to manual segmentation but uses only a fraction of the time of
the latter. The methodology developed in this work can also be applied to
other segmentation tasks in electron tomography or other general areas. It
is especially suitable for applications for which image quality is poor but
there exist strong prior models.
|