Spinal Cord Image Quality Metrics

MRI Processing
1

Hi,

I have several T2 and T2* anatomical spinal cord datasets that I’d like to calculate image quality metrics for (similar to those produced by MRIQC).

I believe metrics like SNR and CNR should be relatively straightforward to calculate; however, I’m struggling to create a good noise mask to use in these calculations. I can create a rudimentary mask by thresholding the images at a low intensity threshold, but I was wondering if there is a more technically sound method (similar to the air/background segmentation available for the brain through SPM).

More generally, if anyone knows of other image quality metrics available for spinal cord data, please let me know. I’m hoping to use them for harmonisation of image-derived phenotypes in multi-site studies (using this tool: https://doi.org/10.1101/2025.06.04.657792).

Thanks in advance,
Lachlan

2

Hi @ltstrike,

Computing SNR and CNR is not straightforward as it can be confounded by multiple factors, including the location of the mask, the assumptions about the homogeneity of the region, B1+ and B1- inhomogeneity, and the algorithm used to combine signal from multiple array coil. Check out this article that describes all the issues, and propose solution (incl. link to code on GitHub): https://onlinelibrary.wiley.com/doi/10.1002/mrm.29249.

Julien

3

Thanks @jcohenadad - that’s exactly what I’m after.

Cheers,
Lachlan