Hi Dilek,
I've actually run SAMSEG-lesion on some gadolinium-enhanced scans in the paper. Look at the top row of Fig. 7 ("T1c") and at the "MSseg" dataset. I did not check the performance of the method on such data. We also don't distinguish between enhanced and non-enhanced lesions.
I would probably set --lesion-mask-pattern to "0" if you want to segment most of the lesions, or --lesion-mask-pattern to "1" if you want to segment only the enhanced lesions.
It is usually not advised to analyze segmentations/volumes obtained using different contrasts as you might introduce biases in your analysis.
Let us know how it goes, Stefano
Hello dear all,
I want to use Freesurfer-Samseg for MS lesion segmentation(cross-sectional). I have read the paper, accordingly I want to try the t1-flair combination in terms of the accuracy.
Do you have any experience with T1 images with contrast agent? I couldn't find anything related to those in the paper. Can I use t1 with contrast in a few patients if the patient doesn't have t1 image without contrast agent? i will declare this situation in the results.
Best, Dilek