D.R.Magee at leeds.ac.uk
Wed Jan 13 10:42:53 EST 2016
[Apparently I didn’t send this to the list first time… I’m really interested in others opinions on OME-TIFF too!]
We use Aperio .svs as the native format for writing output images in our digital pathology image analysis system (http://www.medicalimagemanager.com if were advertising ☺). This is a (reasonably) well described tiled tiff format supporting jpeg and jpeg2000 tiles, which is reasonably easy to write with open source libtiff (+openjpeg for jpeg2000). You also need BigTIFF support for larger files (but I can’t really see a way round that one), and it isn’t readable by standard tiff viewers (again if you are talking a tiled format I can’t see a way round that one either). Obviously the main advantage is it is supported by OpenSlide already (as well as Bioformats).
I have no real issues with svs, but I was also thinking of looking at OME-TIFF too (I like the fact the format is openly described, although a bit too flexible maybe). It isn’t widely used in digital pathology though, but is quite widely used in other microscopy areas. I’m interested in other’s opinions if it will ever be a major format in digital microscopy?
One thing about bio-formats that put me off is that it has a tangle of different licences. Some features require a commercial licence for commercial use (Open Microscopy isn’t quite as open as OpenSlide in that respect). It’s also written in Java, and while there are bindings for other languages (e.g. c++ binding in ITK) it does complicate things (i.e. Java needs to be installed an configured) if you aren’t using Java.
From: openslide-users [mailto:openslide-users-bounces+d.r.magee=leeds.ac.uk at lists.andrew.cmu.edu] On Behalf Of Dmitry Fedorov via openslide-users
Sent: 12 January 2016 17:22
To: Andreas Schmid
Cc: openslide-users at lists.andrew.cmu.edu<mailto:openslide-users at lists.andrew.cmu.edu>
Subject: Re: OME-TIFF
You could use libbioimage <https://bitbucket.org/dimin/bioimageconvert> alongside OpenSlide to achieve full support for OME-TIFF and WSI. In fact it supports pyramidal and tiled OME-TIFF and OME-BigTIFF that are not strictly standardised. It also supports other formats not yet available in OpenSlide, like tiled JPEG-2000 (this will get a lot better with the next openjpeg library version).
Libbioimage is a complimentary library for OpenSlide since it was designed to support multi-dimensional and multi-channel fluorescence imagery. In our system BisQue (management and analysis system) we use openslide for WSI and libbioimage for OME-TIFF for very large images daily. Finally, I decided to integrate OpenSlide into libbioimage as one of the decoders and this should be finished for the next version. I also hope to incorporate OpenSlide extensions like DICOM-WSI from Mathieu Malaterre and have started writing a CZI importer supporting fluorescence.
Although I use OME-TIFF as our main interchange format and do like it, we have encountered a few problems in more than 10 years of use, I'll try to list those for your consideration.
1) Really large images require BigTIFF packaging and thus become unreadable by most programs supporting TIFF.
2) Pyramidal-tiled OME-TIFF is not standardised. I prefer sub-directory based approach (similar to Photoshop's pyramid) myself but implemented support for other cases I know.
3) Every channel plane is stored in a separate directory making it a bit slower to read all the channels since one must change directory several times for every tile.
4) TIFFs with many directories become very slow to read since one must traverse the whole file until the desired directory is reached. I've worked with very large 4D OME-TIFFs with hundreds of thousands of directories. Although libbioimage patched libtiff with a few optimizations for this case it is still a real problem. The solution to this is unfortunately a multi-file solution where each 3D block is stored as a separate OME-TIFF file.
5) OME-TIFF metadata model is quite rigid and designed for a specific use. It does provide a flat name-value list for extended information but it is very verbose and too simple to store a large amount of complex information. Thousands of tags will make that list very slow to parse.
Folks at OME consortium published a paper proposing to use an HDF-5 based format for lifesciences. Although it sounds like a great idea there's no standard definition for it yet. Many companies are using HDF5 based formats for a long time but they are all proprietary, for example Imaris ims5.
On Tue, Jan 12, 2016 at 2:46 AM, Andreas Schmid via openslide-users <openslide-users at lists.andrew.cmu.edu<mailto:openslide-users at lists.andrew.cmu.edu>> wrote:
Thoughts on OME-TIFF as an OpenSlide Format
Actually our company saves the WSI images as an zipped deepzoom pyramid simular to Martin Weihrauchs Smart Zoom Image approach. Unfortunately this format lacks third party support and likely ever will and suffers from the detriments of a non-image format (Like viewing it in any other tiff supported software when it is not too large).
We took some research in the past weeks and considering to switch to a tiff-based WSI format but we are not strigency interested to do our's own thing to be just another one who sheers off from a standardized WSI format.
So we kept an wary eye on the OME-TIFF format.
What do you think? Is OME-TIFF a canditate for beeing an openslide format too?
How do you think will OME develop in the future?
openslide-users mailing list
openslide-users at lists.andrew.cmu.edu<mailto:openslide-users at lists.andrew.cmu.edu>
Dmitry Fedorov Levit <dima at dimin.net<mailto:dima at dimin.net>>
Center for Bio-Image Informatics:
Vision Research Lab, Electrical and Computer Engineering:
University of California, Santa Barbara
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the openslide-users