Apply SCT to segment high resolution mouse spinal cord

SCT
1

Hello,
I have a series of 2D high resolution transverse images of mouse spinal cord that together give the 3D structure (attached a few examples). There is very strong grey and white matter contrast. Could the default SCT NN work to segment the “butterfly” shape of the grey matter in such images? In some ways my challenge is made easier by the fact that the spinal cords have been removed from surrounding tissue. If there is “too much resolution” I suspect it could be down sampled to look more like a MRI scan. But perhaps the NN might be too narrowly fit to conventional human imaging methods.

I have thousands of images, so I’m wondering how fast the segmentation throughput is.

Also, if anyone has suggestion about the state of the art deep learning methods for custom training (ideally with a user friendly GUI) for drawing butterflies I’d greatly appreciate it!

Thanks,
Steve
image

2

Hi Steve,

These are stunning images! We have developed a couple of segmentation models for mouse spinal cord data. Would you be able to share 1-2 representative data where we can test these algorithms, and then point you to the ones that work best?

Cheers
Julien

3

Julien, so glad I remembered to return to this message! My account was new and I didn’t realize I wasn’t getting email notifications. I am very interested in your algorithm. Can we spin off a side chat in direct message? Or is there a way to email you? I’d love to discuss the problem in more detail. Then maybe I can share more of the raw data.

All best,
Steve

4

sure! feel free to email me