Tweet
I am currently on sabbatical at the Institute for Trauma Research and
Biomechanics at the University of Ulm. Here I am working on a project
where we are trying to determine how human fibroblasts align in
response to deformation of the substrate they are growing on. After
the treatment, the cells are fixed and stained. The substrate is
placed on a glass slide where they can be viewed under a microscope.
It is clear that the cells do not develop a single agle of alignment,
but rather, have some angular distribution function. We can quantate
this using an image analysis system, but this requires that a manual
image be made where each cell is replaced by a line that defines its
unique orientation. Althougth the image analysis system is very
sophisticated, I am inexperienced and have not found a reliable
method that is fully automated. Ultimately, I must manually interact
with the image to separate overlapping cells, remove dirt objects, etc.
This is an awkward process to do manually, so I have taken to making
the manual image as a preprocessing step.
I have two question:
1. Is any one aware of techniques that might eliminate the
need for making the manual image or doing a lot of interactive
image touchup?
2. Can anyone give me guidance on how the angular distribution
information can be extracted from a FFT of the image? I am
looking to determine the percentage of angular information
in descrete angle intervals?
Ed Grood
P.S. The silence has been deafening. Our greatest tribute to Herman
is to keep this an active list. I suggest a subject heading of
"Cell Alignment Image Analysis" and public responses. If
some subscribers are not interested they can easily delete any
responses.