IDR-NGFF-2020-09-30

--- tags: NGFF --- # IDR-NGFF-2020-09-30 Attending: Petr, Josh, Mark, Dom, Seb, Frances, Will, Simon, J-m, David - Continuation from https://github.com/ome/meeting-minutes/pull/1 - Datasets - [idr0033](https://idr.openmicroscopy.org/webclient/?show=screen-1751) v [idr0036](https://idr.openmicroscopy.org/webclient/?show=screen-1952) v ... - JRS: [idr0078](https://idr.openmicroscopy.org/webclient/?show=screen-2502) - Especially screenB, nice opportunity for segmentation (I believe the lab has ROIs, but very likely easier to do them ourselves). - Back story-- in each case, the submitting lab will have lots of analyses and it would be interesting if we could give them a “better way” of accessing their own data <-- pipe dream. There may be another one from HPA-- checking and will fill in here. - Seb: have projected images and CellProfiler features - JRS: all the labs _do_ have ROIs even if we don't. - Seb: email them for their ROIs? - Will: All HCS. You want to do some _new_ analysis? With napari etc. you want to load a "pretty" image rather than large screens. - JRS: could add COVID & McDole - Will: idr0077 plants. Seb: candidate for a multi-angle spec - Features - Thumbnail microservice in IDR with updates to read smallest pyramid level - HCS metadata layout/spec for OME-Zarr with updates to omero-cli-zarr and bioformats2raw - Notebooks with a parallelizable analysis - Some form of visualization of the plate layout (javascript, notebook, etc) - Export of the images to OME-Zarr - ZarrReader to re-import the exported images - Stretch goals? - Reading ROIs out-of-band - JSON representation of metadata for SEO - Apply above to idr0080 - Discussion - JMB: currently only napari. updating our clients for viewing. - limit to import from S3 into OMERO. - Josh: and then access S3 on client? Then need renderer client-side. Alternatively, could let RenderingEngine talk to S3. - JMB: prepare for that direction - Josh: as long as we have ZarrReader then we should have the base requirements for IDR. - Will: - will need work on the low-level openlayers data loading. Currently only pass a URL for tile-loading. Might be able to teach OpenLayers to use a canvas as a source for pixel data. - JMB: want to upload "zarr ROI" and overlay on top to see a comparison - Simon: do we stay with iviewer or a new viewer for zarr - Will: all our rendering controls are tied to the server API. Would be very different. - i.e. "Open as" strategy. - Simon: cf. https://github.com/hms-dbmi/vizarr - J-m: may require unifying thumbnails from the secondary client. - Josh: including omero-ms-thumnail might help with that issue if it is *also* detecting - Simon: pick a dataset first? which one would be most popular? - Frances: the differences are: 33/78 are yeast genes and 36 are compounds. Wasn't easy to enrich 78. - Will: email all and whoever replies first? - Seb: all three have related web resources (BBBC as archive & CellVision as atlas). idr0078 has another paper they want to support. Optimistically it may be possible to work with them. - Seb: like 33 & 36 as community reference. Do we know which one is most used? Frances: also idr0080 coming up which is similar to idr0033. - Seb: could also see addressing **Image-Image-Link**, showing the processed-image relationship - JMB: similar to labeled image. could imagine overlaying similar to what figure does. - Thumbnails - walk-through of the microservice logic - Will: worried that there's more work happening on the client - Simon: can also have caching, and having 50+ microservices - JMB: need to consider the non-IDR use-case as well. - Seb: something in the spec which prevents having a non-uniform scaling - Mark: I've wondered if, if we have fairly large thumbnails stored, we could quickly do small-scale downsampling on the fly where it makes sense. - Simon: would get some benchmarking. Seb: in HCS a benefit is that they are all the same. - Summary: - Review [idr0033](https://idr.openmicroscopy.org/webclient/?show=screen-1751) for further discussions. Prepare for extracting the ROIs. - Look into http://github.com/glencoesoftware/omero-ms-thumbnail (Docker exists) to benchmark performance - Start thinking about how to improve ROI handling in the clients. Done: 10:31