2012-12-11 Tuesday Meeting

Dundee: Gus, Jean-Marie, Colin, Helen, Elwood, Kenny, Mark, Blazej, Petr, Will, Jason, Simon, Roger, Chris, Andrew Remote: Andreas, Ian, Ivan (14:41), Liza, Mark, Sébastien, Stick, Josh, Melissa, Douglas, Yuri

Start - 14:36 1. Accepting minutes from last meeting * Done

1. Ian Munro (Imperial) - updates on FLIM analysis (25 min) - 14:37

   • See FLIM OMERO report
   • Ian: Summarise where we are so far.
   • Ian: Started with crisis in software, Matlab in front of C++ processing core, worked with data read from disk. Bottom left corner and raw disk. Still a WIP and being developed. Working with variety of data formats, 2 propriety, LaVision, Becker & Hickl, internal format, tiffs, filename information.
   • Ian: 1st step was to take it and go cross platform, view to integrate with OMERO & add future dist paths.
   • Ian: Work with OMERO formats.
   • Ian: easy 2 formats was first goal, already imported in to OMERO, .SDT files. B&H via bioformats, extra time dim in channels. LaVision convert format to OME-Tiff can be imported, extra time dim in Z. Added capability to laod directly from server. Code is OMERO client.
   • Ian: Write results of processing to OMERO, multi channel floating point image.
   • Ian: Left us with some holes, capabilities on disk that we can’t do in OMERO. Time gated data issue, well plates (multiple images).
   • Ian: In order to do that needed a way to get from disk to OMERO, written own importer.
   • Ian: Where it goes from now, how long it’s required is an open question. Give us something to test code with. Standardised way of storing extra time dimension. The other 2 formats for hist reasons are encoded in c & z, Nice to be able to process extra info as florescence data instead of everything being a special case.
   • Ian: Importer imported from disk to image in OMERO or well / plate.
   • Yuri: Load 1 or many images.
   • Ian: Coming to end of work, optimisation / debugging, matching from OMERO or disk for our application.
   • Ian: Give ability to pick up process part way through. Open stored result in fp format OMERO and visualise using our software.
   • Yuri: Other features, disk interface higher (refs?) for multi channels. Done as an annotation at the moment to a dataset.
   • Yuri: Background connections that need auxillary objects, load them from image file stored in same dataset. Higher ref series of images, aquisition of several 100 of images.
   • Yuri: Projects & Dataset to fit it and save result to another dataset. Scree add new plate so can be scored and stored as new result plate.
   • Ian: Should have floresant lifetime data pretty soon.
   • Jason: Significant achievement, LaVision choosing the multi dim.
   • Ian: Modulo T sems the right way to go, maybe technical reasons for other options.
   • Jean-Marie: Melissa modulo in C @ madison?
   • Melissa: data at madison everything is a channel made sense at the time. No personal objections for doing it in T, want to get antoher opionion.
   • Jason; Contact Curtis @ Madison and make a decision so you can make progress.
   • Ian: 3 formats is okay for now.
   • Jason: Good but need to make a decision.
   • Ian: Might take time for LaVision to adopt solution.
   • Roger: Subchannel in Scifio,
   • Ian: fp data in omero doesn’t look great, understood problem?
   • Jean-Marie: yep, try to get around to it.
   • Yuri: Time thing to load? Pixel wide sweep, some pixels fails.
   • Jean-Marie: Can get a file and have a look afterwards.
   • Ian: Not stopping us, nice if it did display stuff. Not a big deal.
   • Yuri: We have our software for visualisation specifically for FLIM.
   • Jean-Marie: Will contact and get a time frame.
   • Ian: We have our own importer to give something to test with, in the future envisaged, any new poeple in to time imaging will probably start with own format. Way of importing tiffs / pngs. How would someone else newin the field fit in?
   • Jean-Marie: How close to roll out?
   • Ian: happy to roll it out, whether it’s publishable, licences okay biggest problem there.
   • Jean-Marie: Someone else from institute might be interested and provide testing. Organise call later in the week.
   • Ian: Curtis: don’t store data in OMERO for FLIM data.
   • Jason: Writing OME-TIFF
   • Ian: Modulo along t, units are seconds, dont understand pros/cons of modulo solution. Let us know

2. Ivan Cao-Berg (CMU) - OMERO.searcher (25 min)

   • Ivan: Own flavour of OMERO. Content based image search. Search on content / features rather than context. Already have a paper that performs 3 additional tasks, modified OMERO.web interface of the gateway can access functionality. Built django app that users can upload images can compare with features on OMERO database, built local client, communicate with database, return content database, perorma local search against local images based on content of OMERO database.
   • Ivan: extended OME XML for lucene index for searching. Main problem feature calulation can be expensive, system would hang. Feature calculation on demand not anymore, triggers a job to a queue manager then feature calc occurs in the background and appear as features get calculated. Out of box works well. Problem, jpbs sent are heard as OMERO python scripts, username & password is hardcoded in every script, security problem. Last dev meeting, want ability of triggering jobs, without sending info directly. Trivial or not?
   • Ivan: table for features / feature sets, pyton package so people can calculate features and link from scripts, not neccessary from importer.
   • Ivan: feture sets web UI, select an Image, returns images based on feature space. User can decide which queries are desired or not, further improve the results page. In terms of OMERO, worried about security issue on importing, and worried about integration in to OMERO.searcher how can we facilitate that integration. tried as much as possible to not hack OMERO, tried to use OMERO API as much as possible and follow the design. Made it in to a publishable product, how can we improve OMERO.searcher so it looks more like an integrated OMERO product.
   • Ivan: Django app for external search is simple doesn’t follow, 1 image upload, multi image upload, paltes / images/ projects and compare against exisiting images in OMERO. Security issue again, not linking images as OMERO datastructure, uploading a temp image then searching database. Acceptable or not?
   • Ivan: when user queries the database, not making a connection to OMERO, the db just uses the web server.
   • Ivan: Next stage working on is another project. Cell Organiser, main aim is to rain mdoels on raw data and use stat models to generate new examples. Want to test cell organiser in OMERO. OMERO contains the raw data, training of the models in background via scripts and new samples can be linked. Generate and link to OMERO database. Think we have everything we need to perform that task, python version of Cell Organiser (josh) current version is MATLAB.
   • Ivan: Cleaner, more organised code in Python.
   • Ivan: train the model, (bad audio can’t hear :()
   • Ivan: Cell Organiser can produce multiple outputs, that is what will be linked in OMERO, images will be linked, can export, images as files to be imported in Blender, lot of options. Can link stuff automatically.
   • Jason: Not alone with Job scheduling requirement.
   • Jason: Security a blocker?
   • Ivan: Not really a blocker more a concern.
   • Will: Discuss web hooks.
   • Jason: Time to get a meeting.
   • Simon: Setup github yes, send on ome-devel list.
   • Jason: Copies of presentation that we can store? Helpful for writing reports.

3. ASCB Update (5 min) - 15:35

   • Colin: t-shirts delivery tracking
   • Colin: 1 extra laptop in addition / Melissa or Lisa, slideshow or web
     • Liza: fine with mine.
   • Colin: UK/US adapters, airport & US people
   • Colin: Carrying what, distributing things in Dundee tomorrow. All sorted otherwise. Thanks Wilma.

4. Plate import on Gretzky not working - see #9939 - can we rethink this ?

   • Done

5. Glencoe Update - 15:38

   • Chris: Liza coming, jason & Chris have meetings set up, Demoing beta version of JCB data viewer, Carlos & Liza finishing that.
   • Chris: Trying to keep it going during the meeting. Carlos away Monday onwards.

6. Any other business (<5 mins) - 15:40

   • Jason: EMDB has to stay up, emdb.openmicroscopy.org.uk. Mage problems.
   • Kenny: Has ticket in for VM with CLS.
   • Kenny: EasyDNS for ome.org from Stick.
   • Next week meeting, Chris not here for Skype. Get a few more people on Skype, Kenny chase people up (dev contact list).
   • 8hr behind - Time difference

7. Done - 15:42