search

rc / gensei-migrated

Automatically exported from code.google.com/p/gensei
0 stars 0 forks source link

the user could enable or disable the elimination of "boundary effect" #10

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
Currently the objects do not touch the borders of the images in either
direction. In other words, the boudary effect is eliminated. For some
applications (e.g., for application of unbiased counting frame,
http://www.stereology.info/counting-frame/) it would be more "natural" if
the objects would not be enveloped by the background completely, but if
they would "trespass" the borders in all directions. The user would be
allow to choose.

Original issue reported on code.google.com by zto...@gmail.com on 5 Jun 2009 at 11:17

GoogleCodeExporter commented 9 years ago 
Would saving images only from a sub-box of the full box do what you want? The
statistics would be relative to the full box though.

Original comment by robert.c...@gmail.com on 15 Jun 2009 at 6:18

GoogleCodeExporter commented 9 years ago 
No, it wouldn't. By eliminating the boundary effect I meant the situtation, 
where the
ellipsoids are not allowed to reach or to exceed the borders of the full box = 
this
is the current state of the software, isn't it?

But imagine that some real ellipsoids already existed before being sectioned 
and that
the one who performs the sectioning procedure has no information about position 
and
orientation of the ellipsoids. When he treies to make sections, he will surely 
sut
the ellipsoids even with the firts or the last section of the series. This is 
exactly
what happens nearly in all histological examinations and what I wanted to 
simulate. 

There are two files attached ilustrating the issue in 2D. In one of the images, 
the
objects touching the edgeewre excluded from the quantification, i.e. we do not
consider them. In the second image, all the objects (even the incomplete ones) 
were
included into the quantification. 

The idea was to implement a switch, which would be used to:
a) generate only such objects, which do not reach the borders of the box (i.e. 
the
objects are forbidden to reach the borders)
b) generate objects irrespectively to the borders, but include into statistics 
and
images only intersections of the objects situated inside the box or touching the
borders of it.

Choosing b), a problem arises - which objects cut by the borders of the box 
shall be
then counted as intrinsic and which of them shall be not regarded? The best 
solution
would be probably to implement the rules of the unbiased brick, see:
http://www.stereology.info/optical-disector-unbiased-brick/ The rules would be
applied e.g. to the geometrical centre of the ellipsoid. 

I know it is quite complicated, but this would change generation of the objects 
into
a more powerful technique which would resemble the real counting of microscopic
objects in series of sections (numerical density of objects).

Original comment by zto...@gmail.com on 16 Jun 2009 at 2:21

Attachments:

GoogleCodeExporter commented 9 years ago 
If there would be a problem to summarize statistical data based on analytic 
formulas
(when the first/last sections would contain any structures), the approximative
results of volumes and surfaces would be enough. 

Another solution suggested before was to create a "super-box" surrounding the 
actual
"subbox" with objects intersecting the boundaries of the subbox, but not those 
of the
super-box.

Original comment by zto...@gmail.com on 4 Sep 2009 at 9:51