[ome-devel] OMERO/Matlab

dzmacdonald donald at lifesci.dundee.ac.uk
Fri Dec 18 16:32:59 GMT 2009


Hi Jonas,
On Dec 16, 2009, at 3:19 AM, Jonas Dorn wrote:

> Thanks for your replies, Michael and Donald.
>
> I have finally managed to find the OmeroMatlab documentation on the  
> site, and I am currently trying to install omero.server on my laptop  
> so that I can test some examples. However, looking over the examples  
> on the ome page, I noticed that they are mostly concerned with  
> reading images from the database. This is not what I need most.
>
> I think I should try and clarify some more what I'd like my workflow  
> to look like:
>
> Image analysis
> - Find experiments or images that have not been analyzed yet  
> (calling the database from Matlab)
You could add an image annotation/tag to say that it's been analysed
> - Pass filenames of images to Matlab (I have implemented my own  
> readers in Matlab to read the pixels directly from disk, which is  
> most likely faster than accessing images via the database interface)
This can be done by getting the image object and reading it's name.
> - Run analysis in Matlab
> - Attach ROIs to images as soon as they become available in the  
> analysis (ROIs should be stored in the database so that I can access  
> them from multiple computers, correct?)
They are, you can create an ROI and it will have an image id  
associated wit it.
> - Annotate ROIs and images as soon as the information becomes  
> available in the analysis. This annotation includes status of the  
> analysis (when has what analysis step been executed), but also some  
> results, for example loading speed.
> - Attach graphs to ROIs.
> - Attach path to analysis files to images.
>
The structured annotations will allow you to do all of the above.
> Data browsing
> - In Omero.insight, find images/ROIs with certain characteristics
> - Look through attached graphs
> - Improve/Extend annotation, e.g. to classify ROIs according to  
> phenotype.
>
Yes that's all possible, insight allows you to search for annotations.

> Data analysis
> - Find images with completed analysis/annotation with specific  
> characteristics by querying database (calling the database from  
> Matlab or from insight)
> - Run statistical comparisons between sets, display group averages/ 
> distributions using Matlab scripts, run more involved analysis in  
> Matlab.
>
> In short, I would like to use OMERO not so much to be able to  
> interact with the image data, but to have an easy way of accessing  
> subsets of my data with specific properties, be it analysis results  
> or processing history.
>
Yes that's all possible via the iQuery interface and structured  
annotations.

> To this end, I need to
> 1. Populate the database with references to images - all that OMERO  
> needs to store is the location of the raw image file. The pixels  
> themselves should not be imported into the database, because I have  
> barely enough space for one copy of my data, thus an import that  
> copies data into OME-TIFF is not feasible (Michael, does this answer  
> your question?).
> I do not plan to touch the raw data file at all, thus I would have  
> no problems accepting that the raw data becomes read-only upon  
> import. Also, I do not plan to move the data (again, there are not  
> that many shares with enough space on the server).
> Donald, for my project all images are deltavision files already, so  
> it looks like your idea should be no problem to implement.
Great, adding new images to omero is easy, just create an image object  
and use the
updateService to add it to the DB.
> 2. A way to attach ROIs to the images. I currently store ROIs as a  
> fixed rectangle plus, for each time point, a mask within the  
> rectangle. I guess that OMERO does not accept arbitrarily shaped  
> ROIs (maybe I just didn't quite understand the documentation,  
> sorry). I think it should be fine to just pass the rectangle to  
> OMERO. However, are ROIs allowed to overlap?
OMERO allows the storage of Mask ROI. ROI's can overlap. There is no  
restriction on ROI, they can span z, and t, and be made of arbitrary  
shapes.
> 3. A way to attach processing information and (scalar) analysis  
> results to ROIs and images. I was first looking at  
> StructuredAnnotations, but it is, of course, a problem if their  
> values are immutable - I need to change their value at least once.  
> Ideally, I could annotate the data in a form of 'parameterName':  
> parameterValue, though it seems like misusing the protocol editor  
> for this may be the only way to get this to work - can I set the  
> 'experimental parameters' from Matlab?. Tables also look like a good  
> possibility, though I would prefer not to mix processing information  
> and analysis results in the same table. Is thus possible to attach  
> multiple tables to ROIs? Also, is a table editable in OMERO.insight?
No you cannot edit a table in insight, but you could use the table  
structures to hold the results,
> 4. A way to query these annotations from Matlab or from OMERO.insight.
That's all possible, though you can only query structured annotations  
in isight, not the tables.
> 5. A way to browse attached figures in OMERO.insight (so that I can  
> use the information presented by the graphs to manually annotate the  
> ROIs.
>
You can only view roi from insight, the only annotation that can be  
made to the roi in insight at the moment is text on the roi.
> Can I do all this with the current version of OMERO? What would I  
> need to change? Could you help me with general examples on how to  
> accomplish these steps from Matlab?

most of this can be done inside OMERO, but I think that the structures  
we're planning to write for scripts will help with some of the more  
complex workflows, and hopefully remove the need for using structured  
annotations.

Regards

D.
> Sorry for the wall of text and thanks for your insights!
>
> Jonas
> _______________________________________________
> ome-devel mailing list
> ome-devel at lists.openmicroscopy.org.uk
> http://lists.openmicroscopy.org.uk/mailman/listinfo/ome-devel



More information about the ome-devel mailing list