[ome-devel] 3D data

[Mark Carroll]

On 03/12/2018 10:11 AM, Eilidh Troup wrote:

> I have a question from a potential OMERO user about 3D image data. How well would OMERO handle the following data?
>
> We have large imaging datasets from our 3D plant phenotyping system. Not microscopy images but whole plants (growth over time). Would these fit with the OMERO database remit?  Our data format is .bmp files, a specialised analysis file for the 3D data (.npz), and .properties file.

There are various ways of combining 2D BMPs into time series that can
then be scrolled through in OMERO.web and laid out in OMERO.figure. For
example, there is the Combine Images script mentioned at
http://help.openmicroscopy.org/scripts.html#utility and even at import
time pattern files can be used as on
https://docs.openmicroscopy.org/latest/bio-formats5.8/formats/pattern-file.html.

I would expect OMERO.tables at
https://docs.openmicroscopy.org/latest/omero5.4/developers/Tables.html
to fit nicely with the npz files and the properties files could become
map annotations as at
https://docs.openmicroscopy.org/latest/omero5.4/developers/Model/KeyValuePairs.html.
However, one would need to write a simple script, perhaps using OMERO's
API and/or gateways, to perform those conversions. Of course, arbitrary
files may be directly attached to imaging data in OMERO.

In short: OMERO could probably handle the data well but to make it
usefully viewable and queryable it would probably help to at least write
a bit of Python code to slightly reprocess the data into something more
OMERO-friendly, whether before or after import. OMERO offers a few
options for exactly how that happens: for example, some of the
properties file data may most appropriately be represented as tags in
OMERO if some keys take few possible values.

I can't resist speculating that for your application regions of interest
could be valuable too, perhaps via our MATLAB or ImageJ integration: for
storing quantitative measurements from each image about how the plants
are growing, then for subsequent querying and analysis over the time
series. If the numpy arrays do include points or shapes then it could be
worth writing a script to process, say, some of that npz data into OMERO
ROIs.

If your data format is or may become widely adopted then it could be a
good investment for you to write a Bio-Formats reader that turns the
data into map annotations and ROIs at import time but I am guessing that
some ad hoc client scripts are your best path forward for now.

Cheers,

Mark

The University of Dundee is a registered Scottish Charity, No: SC015096