Can msseg be used to calculate white matter lesion load?

[johnscot]

Hi,
I am looking to calculate white matter lesion load i.e the total volume of white matter hyperintensities. Can MSSEG be used for this or would you recommend another method ? I have been exploring the use of the LST toolbox.

Using the output from MSSEG, I took the 'image'_refilled_candidates image and ran fslstats -V. My feeling is that this should be the white matter lesion volume? Do you think this is a correct approach or is the purpose of MSSEG really just to fill lesions to improve the robustness of GM, WM and CSF volumes?

Many thanks and any advice appreciated
John

[sergivalverde]

Hi,
The main problem trying to find WM lesions with MSSEG is that is designed to remove only those that are hypo-intense in T1-w. So, for having the estimated lesion mask in T1-w, just use:

options.debug  = 1

Then, after segmentation you should have a folder .run with different intermediate files, one of those being the estimated lesion mask : )

Anyway, I you want some unsupervised classification, I would definitely use other options as LST or our SLS method. SLS is also Matlab based and showed some benefits over other state-of-the-art methods (see paper).

If you have a reasonable number of manually annotated lesion masks, I would use this . This is Python based and maybe still tricky, but the performance is significantly superior. (see paper)

[johnscot]

Thank you. I'll look into your recommendations.
Best wishes
John