[ome-devel] annotations

[Bernd Jagla]

Tom,

|
|> Some of the issues I see are that the only reliable way to link the
|> data is through the directory structure. The spreadsheet importer
|> needs images IDs that have to be obtained from OME so a simple script
|> to re could help you expandformat the results file wouldn't work.
|Actually the spreadsheet importer works on names (OME names) too.
|The spreadsheet importer can do alot more than the web documentation
|(it's out of date). Try ome help annotate wizards for more info on what
|is possible. We can talk off the ome-devel list about this.

The spreadsheet importer has no chance in knowing for sure which images the
values belong to. Since I don't want to write a script to update and
reformat the text file, I don't consider this an option at the moment.

|
|Please take a look at the spreadsheet importer again. It is really
|powerful and can get alot done. I am not 100% sure about what it is you
|need to do, but it might be the case that the easiest way to proceed is
|to expand the the importer`s functionality. I can do that for you.
|
|One thing I was wondering about was how to make the spread-sheet
|importer also use SHA1 digests or original-file names (not just OME
|names/image ids) as foreign keys. I didn't implement it because we
|didn't have a good use-case. You might be just the use-case I was
|looking for.
|
|I think the right way of thinking about your problem is: "can I write a
|script to convert the results file into a csv file were each row is an
|image and the columns are image annotation information". You can do some
|of if programatically, and tweak the rest in excel.

I think I will implement the import of the analysis files within the
getGroups function of the importer. Here is what I am thinking of in terms
of algorithm:

getGroups will have to do most of the work. It will read in all the analysis
files that are under the specific directory structure. A readAnaHeader
function will read in the parameters of the analysis and maybe generate a
new ST and return the corresponding mexes.

=> how can I create a new ST from within a Perl routine? 
=> How do I erase an ST? (Since I probably won't do it right the first time
and I don't want to reload the database for each trial, I need a mechanism
to update the ST from either the command line or a small script to remove
it).

There will be a different ST for the different analysis modules the INCell
provides (Analysis modules are functions that applied to the images, given a
set of parameters/thresholds and returning different kinds of real values).
The number of these modules is changing all the time and I would like to
avoid creating new ones manually each time GE comes out with a new module.

Then, the values for all the parameters/thresholds will be stored in an
instance of this ST. Those ST's should be global.
The results will point to the module instance. For each of the different
module ST's we also need an ST to store the individual values and links to
that module-ST and the image. (Here values describe the specific thresholds
of the analysis module and results describe the values derived from the
images when applying the above mentioned algorithms.)

getGroups will also read in the results and put them together with all the
needed Mexes in the "groups" structure that is returned by getGroups. Since
the image Mex is not known at that time we have to populate the tables after
the images are read in. 

I will have to check if the results have already been written to the tables.

Also, I will check if the images have been read already since uncompressing
them is really time consuming.

Does anyone know how to go about entering new attributes when some of the
parameters are not known at compile time? For example: The following
construction will generate a new Plate instance:
				my $plate_attr =
$factory->newAttribute('Plate', undef, $global_mex,{
					Name => $plateName,
					ExternalReference => '2 B named'
				});

But for my purposes "Name" and "ExternalReference" will not be know, nor
will I know how many of these parameters will there be. This information is
all hidden in the analysis files....

Please let me know if you have any concerns or suggestions with this
approach.

Thanks a lot

Bernd