SVS Images on CellProfiler

Dear CP2.0 Help,
Does CellProfiler support .svs images (such as those produced from whole slide scanners from Aperio: loci.wisc.edu/bio-formats-fo … o-svs-tiff)?

Here is an example file (they tend to be huge – 500 -1000 mb):
tcga-data.nci.nih.gov/tcgafiles/ … 00-DX7.svs

It seems there might be support for svs images in cellProfiler through BioFormats? (svn.broadinstitute.org/CellProf … init_.py).

Is it possible to directly read svs images into a cp pipeline?

Thanks,
Andy

Hi Andy,

CP could be updated to include .svs support since, as you noticed, Bio-Formats supports it. However, since you also mention that the images are extremely large, you should note that even if support was enabled, you would be subject to the memory constraints mentioned here.

Regards,
-Mark

Hi Mark,
Thanks for the information! It seems these svs’s will be too big to read directly in CellProfiler. The best solution will probably be to convert SVS -> TIFF, and then break the TIFF into non-overlapping TIFF tiles prior to reading into CellProfiler. Although this is not directly related to CellProfiler, do you have any experience with converting SVS to TIFF?
Thanks very much!
–Andy

Hi Mark,

We tried using Bioformats to convert our SVS files to TIFF, but it returned an error saying that the images were too big. Do you happen to know of any alternative software?

Thanks!
Pang Wei

Hi Andy and Pang,

We have not worked with these files ourselves, but some Googling turned up this ImageJ forum post; apparently others have had the same problem. Some further Google investigating may turn up other helpful links.

Hope this helps,
-Mark

Hi,

I break up SVS files using Python scripts that call Pyopenslide which are Python bindings for Openslide. I then use CellProfiler to analyse these “sub images”. CP’s good for stuff like identifying nuclei, but sometimes the fields you’d like to analyse are just too big for CP, for example “Identify Primary Objects” has a limit of IIRC 999 pixels.

There’s no hard limit to the size of images that CellProfiler can handle and 1000x1000 images should be comfortably handled by a typical desktop computer. However, you are right - CellProfiler is at this point not optimized for very large images and performance will degrade precipitously once the memory required for an image set exceeds the amount available.

Dear Andy,

There are two issues you’re dealing with right now; file size and file format.

To deal with file size, I would recommend either tiling your image into subimages or selecting a subset of fields to analyze. Remember that if you’re using very large images, you’ll need either a lot of time or a lot of computing power to analyze them.

As for the .svs file format issue, there are a few possible options:

  1. Use Aperio ImageScope (aperio.com: free download) to Extract Region (the button next to the pen tool). This gives you the option of extracting to a single large TIFF, extracting to tiled jpeg images (CWS), or others. Remember that TIFF images have a maximum image size of 4GB (compression won’t help) and that 32 bit operating systems can’t open images larger than 2GB.
  2. Contact Aperio and obtain a copy of either Digital Slide Studio or the Icomp.exe command line tool. These tools should be installed on systems with Aperio Spectrum.
  3. Utilize the other methods described below.

Good luck!

Jesse

[quote=“ahbeck”]Dear CP2.0 Help,
Does CellProfiler support .svs images (such as those produced from whole slide scanners from Aperio: loci.wisc.edu/bio-formats-fo … o-svs-tiff)?

Here is an example file (they tend to be huge – 500 -1000 mb):
tcga-data.nci.nih.gov/tcgafiles/ … 00-DX7.svs

It seems there might be support for svs images in cellProfiler through BioFormats? (svn.broadinstitute.org/CellProf … init_.py).

Is it possible to directly read svs images into a cp pipeline?

Thanks,
Andy[/quote]