[ome-users] user feedback part II

[Chris Allan]

On Mon, Jul 04, 2005 at 11:47:11AM +0100, Alastair Kerr wrote:
> More Questions about OME use
> 
> 23/6/5
> 
> We eventually got round to adding one of our previous queries into
> bugzilla as requested (sorry for the wait, a confusion of responsibility
> at this end). Here are some more user feedback. I hope this is useful to
> you all.  
> - Alastair 
> 

Hi Alastair,

I'll respond to what I can and leave the rest to the others.

> 
> Specific questions about use (OME 2.4.0)
> 
> Find Spots
> we failed in our attempts to run the ???find spots??? analysis (Web UI) ???
> any ideas?
> Error Message 
> 'There was an error executing this chain: Can't method
> "attribute_sort_time" on an undefined value at
> /usr/lib/perl5/site_perl/5.8.3/OME/Analysis/Engine.pm line 852'Note 
> 
> Note that the user had difficulty understanding the functionality. 
> Perhaps this could be solved by additional documentation?
> 
> 
> ROI
> (update on previous question about ROI) Using the ROI definition tool
> (shoola) ??? first open a single channel data set then set up an ROI and
> apply. Slightly odd in the way it works: don???t get ROI analysis data
> (analysis context) to work until you first go into ???rendering??? and
> select greymodel (does not select this as the default) when tried on
> MAC. With PC seems to recognise a channel and enter it but not do
> anything until you click over the select channels and select Roi???s so
> that they are both highlighted blue.
> 
> 
> Import/Export of OME XML problems. 
> An image file was exported from OME as OME XML format and then import
> was attempted with the same filename. No error message but imported 0
> images from 0 files.  Similar with a filename not already present in OME
> but the OME XML file format generated by Imaris. Imaris did not import
> an OME XML file outputted from OME but would reimport an OME XML file
> that it generated itself from a Deltavision file format.
> Error message from Imaris 4: ??? (file type not recognised) ???
> 
> 
> ImageJ (setup with OME plugin with assistance from Curtis Rueden) had no
> problem directly importing from OME, modifying the pixel data and
> exporting back to OME as another pixel set under the same filename. What
> should happen if you upload a file originally exported from OME to the
> same dataset and image name? We were looking at this as a way to add
> additional pixel sets to an existing data file. We have also struggled
> with ???create a custom annotation??? ???pixels??? but it did not seem to add
> the selected file data as a Pixel Set.
> 

Right now a significant amount number of our tools -- certainly the web
interface's viewer and Shoola's viewer -- are not aware of multiple
pixels sets. They're only setup to view the "default" pixels set. We
have some data model things to work out as well before this
functionality is uniform throughout our software.

In short: you can add additional pixels sets to an image but right now
we don't let you view them.

> 
> 
> InstrumentsAnnotation
> - When an instrument semantic type has been defined we still are unable
> to return any hits when we search for ???instrument??? but we can find hits
> under ImageInstrument where we have linked image files to a defined
> instrument as a custom annotation. If you then search again under
> Imageinstrument it fails because it leave @Instrument in the ???type??? box.
> Perhaps it is because we are currently putting in the instruments and
> want to check which ones have been done, but it seems odd that you
> cannot find instruments entered into OME unless they have been linked to
> an image file. (is this true?)
> 
> 
> 
> 
> General Questions about use:
> - We are still having problems with Safari on the MAC and explorer on
> PC. On the same machines FireFox works without problems.  We can get to
> the login window but do not progress. We tried fiddling with setting for
> cookies and popups without success. (I went though the list that Ilya
> gave me last time. Given that firefox does work, the problem must be
> client side. Damned if I can find a solution).
>  
> - Would it be fair to say that the best use of OME is (at least
> currently?) with newly generated data inputted immediately after
> collection, rather that for retrospective archiving of stored data from
> CD???s / DVD???s ??? especially where the raw data has been processed to
> generate further pixel sets?

Actually, the best use of OME is after your "batch processing" step or
the step at which you feel the image is ready to be worked with. For
your users I suspect this would be after they have deconvolved the file.

> 
> - We have been looking to see if there is a function that lets you
> specify that an image data set should be entered as an additional pixel
> set on an existing image entry ??? a specific interest is where you have
> raw and deconvolved image data sets as separate files to enter into OME
> and combine under a single image file name. So far we have not been able
> to accomplish this except through ImageJ.  Advice in achieving this
> would be welcome. 

As above the support for this is limited at the moment.

> 
> - we would really like some batch functions which would semi automate
> the data upload process would be great even extending this to
> annotations e.g. :
> *import all files from a particular folder into OME under their
> corresponding filenames ??? ideally with a dataset name taken by default 
> from the folder (with the option to specify a different name).

We can certainly look at doing this from the web interface. You can of
course do this from the command line already but that doesn't help your
users.

> * batch append annotations or other metadata to image sets (again
> possibly as part of the import?) e.g. specify that all images came from
> a particular instrument / objective / 
> EX and EM wavelengths / require a particular OTF???
> 
> - Can chain builder be used to automate import in a batchwise way?
> 
> Related to the above, has anyone (Jason?) set up deconvolution queuing
> that is able to collect batches of image data from OME, pass the
> appropriate file format to the deconvolution algorithm and (after visual
> inspection by the user) return the deconvolved pixel sets appended with
> the deconvolution parameters?  Automation related to deconvolution would
> be a key criteria for our routine use of OME. I might be able to write
> some perl wrapper for this but some sort of perl API or code that I
> could hack would be a big help. 
> 

In Dundee we have this automated deconvolution setup in a testing
environment outside of OME. Deconvolving DeltaVision data is tricky
because of the proprietary nature of the deconvolution algorithms and
their interface. Essentially you're stuck with a command line version of
decon3d which only takes DeltaVision files as input, requires a
significant number of command line options and is non-portable across
different processor architectures.

At this point, unless Applied Precision or another deconvolution
algorithm vendor wants to open up their API or work with us to provide a
cross platform and integrated solution I don't see us providing a
deconvolution solution as part of OME.

If you want to discuss the technical issues of this further you're
welcome to contact me offlist.

> 
> - We have looked at ways of organising data at the time of capture to
> facilitate uptake by OME. Embracing the ideas of simple project/dataset
> organisation and that a dataset should contain data of a similar type we
> have tried file structures where the directories correspond to what will
> be / are projects and datasets. Nested directory structures help
> organise multiple treatments (condition variations) + controls +
> dependent and independent repeat data and experiments, however, they
> make upload into OME laborious.  A simple organisation of ???folder = 
> datasets??? seems easiest if you name the files and folders well. 
>  Using imaginative file naming always seems to break down eventually.
> Any thoughts on naming / incrementing datasets and files would be
> welcome. We have:
> ???user initials_date_celltype identifier_treatment identifier_repeat???.
> Should OME automatically append incremental numbering/ identifying
> lettering to datasets (like softWoRx DV file format appends filenames
> according to incremental image number or processing steps)?
> 
> - It looks like raw data needs to be put onto OME as soon as possible
> after collection, before any processing. Are there any projects
> collaberating with softwre such as metamorph to include data import at
> time of data capture?

I'm not currently aware of anyone trying to integrate Metamorph's data
collection with OME, no. We'd certainly welcome any input on this front.

> 
> We need assistance to find the way to add in missing specific metadata:
> such as
> Missing EX/EM data, and the objective used or which OTF to use/was used.
> (We have started to store OTF images in OME) .
> We found in shoola that you could view a table of image info data but
> could not edit it (right click on preview image). With the image
> displayed and from the Rendering / Channel Editor / Grey Model
> information button there was an information table that could be edited
> but not all of the input was remembered (only the excitation and dye).
> On the image information no emission was specified, however the channel
> was specified as emission 565. (Original data imported into a DV format
> but collected using a custom cube on non-DV instrument ??? header edited
> for deconvolution).
> 
> - We don???t see a way of deleting images from OME but can see that this
> is a possible source of difficulties with incorporation of files into
> multiple datasets potentially owned by different people. How do you go
> about, firstly, dissociating an image from a dataset and deleting it
> from  OMESIS? 
> 
> - In the Web UI the view of search has folder icons that don???t do
> anything ??? could they be clickable to give a tree structure of the
> project/datasets in a similar way to the Data Manager View or Hierarchy
> browser in the Java Client. This would be especially useful if nested
> datasets are included. (From a user who like programs such as file
> explorer in windows). 

Ciao.

-Chris