Aperio slides with 33003 compression

[Benjamin Gilbert]

On 12/12/2012 03:49 PM, Patrick, Todd wrote:
> We have been trying to tile an Aperio slide with 33003 compression
> for use with the NYU Virtual Microscope
> (http://code.google.com/p/virtualmicroscope/) project.  We are using
> the slide tiling python scripts provided with the project.  The
> scripts use OpenSlide to process SVS files using 33003 compression.
> However, we have hit a snag with the larger SVS files we want to
> process.
>
> One of the slides is roughly 98000x42000 in size.  We have run across
> some segmentation faults and values that are too large to be stored
> in certain data types (i.e. int types).

That script asks OpenSlide to load the entirety of the 
highest-resolution slide level into RAM.  Needless to say, that's not a 
good idea.  pyOpenSlide will segfault if malloc() fails, and as you 
discovered, PIL can't handle large images.

These days, VIPS can use OpenSlide to read slide files directly:

     https://github.com/openslide/openslide/wiki/OpenSlideAndVIPS

so you could modify tiftiler.py to do everything through VIPS.  The 
development version of VIPS even knows how to produce Google Maps tiles, 
which might allow you to avoid the script entirely.

> We would be willing to provide some test data, particularly the
> current SVS file that we are having issues with.  Maybe with someone
> else taking a look at this problem slide, these issues could be
> resolved quicker.  Where can I send this SVS file?  Thanks

Thanks for the offer!  I'll contact you privately to discuss.

--Benjamin Gilbert