Importing Leica .SCN WSIs above 2 Gigapixel limit

I am trying to use Bio-Formats to import multi-gigabyte .SCN whole-slide images for segmentation in FIJI. I believe they are “Leica SCN 21”, according to the image information tool when viewing the images in ImageScope. My sample image is 1.4 GB, and is 61,032 x 46,705 x 1 pixels. From my limited understanding, it’s a pyramidal image and the file contains 15 series including a slide label and various macro/thumbnail images at different pixel dimensions (the highest resolution series is the one I want, with dimensions listed above).

I can use Bio-Formats to import the one series I want but receive the following error:

…which I interpret to mean that I am above the 2 gigapixel limit in Java/ImageJ2 based on what I read in other posts. I know that I can use the Crop on Import function to manually load in small portions of the image at a time, but I have hundreds of images to import and segment and this doesn’t seem practical.

Is there a way to load in my whole images at once, maybe with different settings or another plugin? Or is my only option to do some kind of compression or downsampling on all of my images?

Thank you in advance.

Hi @caligos

yes you are hitting one of the limitations of the classical ImageJ and as you note, cropping on import does not scale if you are dealing with lots of whole slide images. What you will likely need is some tool that is able to introspect and navigate through the pyramidal levels of your Leica SCN file rather than attempting to load the entire full resolution.

There are a few open-source and commercial solutions that were designed specifically with this imaging modality in mind discussed in a few forum posts - see for instance Best software to work with large images, Importing a .vsi composite-image in ImageJ via bioformats-plugin - how do I get a high resolution Macro image. From the open-source side, BigDataViewer, QuPath, Orbit are all relevant options and their developers are quite active on this forum.

Hi @s.besson
Thanks for your answer. I guess I’ll have to keep looking (thanks for those links). I have already looked into BDV and Orbit as well, but I haven’t tried QuPath yet.