Posted by
Michael Schmid on
Jul 09, 2009; 8:46am
URL: http://imagej.273.s1.nabble.com/Attachmentpoint-problem-tp3691862p3691865.html
Hi Maringa,
this image looks very different than your schematic drawing some two weeks
ago!
The main problem is identifying the soma. A very simple, but not very good
approach is simply binary "open" with a few iterations (e.g. 4) and the
default count=1. Update to the new ImageJ version 1.43c, then you can have
preview in the binary options and you see immediately what you get.
Run 'outline' on the soma.
Then skeletonize a copy of the input images, shorten the branches a bit
(erode with count=7, iterations e.g. 2) to get rid of noise.
Superimpose the skeletonized image and the soma outlines with the image
calculator 'average'. Threshold the 255 pixels - those should be roughly
where the dendrites attach to the soma. You can apply the selection and
use "find maxima" to get single points.
It is only a rough procedure, you will have to refine it. E.g. fewer
iterations for 'open' followed by the particle analyzer with a range of
circularity and size suited to better identify the soma. Also, replacing
"Open" with n iterations by Process>Filters>Minimum by a radius of n
pixels and then Maximum by the same number of pixels will better preserve
the shapes of the soma than 'open'.
Michael
_____________________________________________________________________
On Thu, July 9, 2009 06:41, maringa wrote:
> Hi Michael,
>
> Here is my binary image:
>
http://img195.imageshack.us/img195/9354/20070510002005000i2max2.jpg>
> Maringa
> ________________________________
>
>
> Hi Maringa,
>
> can you post the binary input image to some server? (if it is too large,
> don't scale it, crop it).
> There are many ways to create the binary input from your color image.
> Without seeing the binary image and trying for myself, I can't determine
> the cause of the problem.
>
> Michael
> _____________________________________________________________________
>
> On Wed, July 8, 2009 09:29, maringa wrote:
>
>> Hi,
>>
>> I posted a question a couple of weeks ago, about how to detect and count
>> the
>> attachmentpoints
>> of a soma (where the neurites attach to the soma) I got the advise to
>> skeletonize two images, and
>> dilate or erode one of them with 1px and then subtract them from
>> eachother.
>> This just generates a lot of dots. I skeletonize, then set "count" in
>> Binary
>> Options to 6 and erode. Which should leave only the crossings. I also
>> combined this by dilating with 'count=1' and 'iterations' a bit more
>> than
>> the line thickness.
>> Then run 'Find Maxima' to get single points or count the patches (former
>> crossings).
>>
>> However, the result is this:
>>
http://img193.imageshack.us/img193/7715/neurph4.jpg>> just a lot of small dots...
>>
>> Does anyone have another sugestion? It's kind of annoying to be stuck
>> whit
>> this simple (?) problem...
>>
>> Thanks in advance :)
>>