[ome-devel] About ROIs...

[Anatole Chessel]

Thank you very much for the quick and comprehensive answer.
>
>> -Rois are not shown on the projections. Obviously showing shapes from 
>> various z on a single plane is not straightforward, but it might be 
>> very handy in some cases (a 2D segmentation propagated to the whole 
>> stack for example).
>
> Interesting use case.
May I advocate for that option to be included if/when you touch the ROI 
GUI? Having a show all/hide all button could be handy too. And I did not 
found how to delete one roi, only either one shape or all rois. (and 
also: is it possible to measure stuff (ie mean fluorescence, etc...) at 
the level of roi, not shapes?)

>>
>> More generally, if I understand correctly, rois are defined as a 
>> collection of 2D x-y shapes each of them having a depth z, a time t 
>> and possibly a channel c. One can imagine cases when this modeling 
>> would not make things easy (3D segmentation, single tracks in time, 
>> tubular objects in 3D...). A related note is hierarchical ROI, 
>> holding for example a segmented cell and intracellular 
>> compartment/objects. But this would be trying to see roi as an 
>> intrinsic way to hold arbitrary geometrical data, which they I guess 
>> were not designed to be. Any thought on this/plan to extend the 
>> current model?
>
> Following various discussions, we made a proposal to add support for 
> ROI 3D.
> http://www.ome-xml.org/wiki/ROI/ProposalSeptember2010
>
> If you can have a look at it and let us know what you think.
>
> The ROI model-DB/clients alignment will happen later in the year.
>

Interesting. A few semi-random comments I can share:

Wouldn't a few well chosen 3D primitive be useful as well? Like, 
sphere/ellipsoid, tube (=line with a 3D width)... And I am guessing 
input/output of VertexArray into standard format for viewing or 
processing (CGAL for computational geometry comes to mind, openGL stuff 
as well I guess) would be useful too, although that may be more for 
scripts/outside clients...

What about shapes intrinsically in 4D or even 5D? It might be a pain to 
deal with but one can think of cases where it may come in handy 
(tracking of stuff merging/splitting for example, or maybe sometime 
having one track=one shape could be useful; for channel: FRET pairs and 
photo-switchable protein, maybe?).

The other point is hierarchical roi: would it be possible to have nested 
roi (rois having some kind of RoiRoisSeq field for example)? it would 
allow for intra-cellular localisation in particular, but most likely 
other uses.

Is there a time-scale for the changes in the model?

Many Thanks...

-- 
Anatole Chessel
Research associate
The Gurdon Institute
University of Cambridge
Tennis Court Road, Cambridge CB2 1QN
tel: +44 (0)1223 767197

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openmicroscopy.org.uk/pipermail/ome-devel/attachments/20110509/89ce2339/attachment.html>