Large Blood Slide Scans

Here is a series of blood slide scans I made using openflexure microscope and fiji grid collection stitching plugin.

http://slides.lat

5 Likes

This is really cool, thanks for sharing! I am curious to learn more about the software you used - were you controlling the microscope from Fiji directly, did micromanager get involved somewhere, or were you able to control it over the HTTP API? If you used the built-in scanning feature and then tiled it with Fiji, that would be good to know too. Funnily enough I was talking last week to someone about how to get control of an OFM from micromanager.

Also, is the website something you’ve put together, or a service that’s more generally available? It looks great!

1 Like

I put together the website and posted it on here a few months ago. I just made this public instance of it where I striped out the ability to save changes and upload new slides. It basically just tiles a very large image to be viewable by the leaflet javascript plugin. Then I added some features to do basic blood slide analysis tools like whats done in my work at a clinical lab.

If anybody has a slide scan they want to share I would be happy to upload it on the public website!

For scanning the image I used ofm python package. I used a custom function because I wanted to download the images as I scanned and because I used 3 tries of auto focusing which I think you can do with the built in scanner but it would mess up the image numbering. I tiled the packages with a macro in pyimagej. Here is my script GitHub - v72nico/OFM-Imagej-Script

2 Likes

Very nice :slight_smile: I’m impressed at how quickly those big PNGs load - I’d have thought you would need to be using something tile-based like SeaDragon, but perhaps that’s only a must-have when the scans get really huge…

Yeah it works pretty well. The total open street map which is commonly rendered with leaflet is terabytes in size, because only the area you request is rendered it can be quite quick.

The large slide on there (slide 74) was never compressed and totaled at 2.1 gb.

1 Like

neat! I was confused by the github repo only having single PNG files, but I guess leaflet must be dividing it up into tiles when the site gets compiled or something - it clearly loads way too fast for me to have downloaded all 2.1Gb!

1 Like

Nah I had to write the code to tile the images. It converts the large images down to 256x256 tiles at multiple zoom levels. Then I feed that into leaflet.

I couldn’t let the packages do all the work for me :wink:

2 Likes

thanks! Future versions of the software might generate those tiles directly (to keep memory footprint down) - at the moment they’re usually combined into a pyramidal TIFF, but it might be that we could add a feature that makes it easily viewable with leaflet (or seadragon).

1 Like

Hi Nico. I am very impressed with your work. Does your code analyze the red cell morphology or is this a manual feature you added?

1 Like

Hey Daniel. All of the red cell morphology and wbc categorization is manual. My current sample is only some slides of my own blood, so I am limited in what I can do.

If I had more samples it would be cool to do something with cellpose to identify rbc morphology and preform wbc categorization.

Hi Nico. I am truly interested in your work. As a pathologist, I have daily access to countless blood samples. I would like to have a chat with you. I am working on a global health project that could use something like this if you are interested.

Sounds interesting! I sent a private message with my contact info.