Dice Score

[apjklab]

Hello authors,

Thank you for your work.

I had a question regarding the TractSeg bundle segmentation pipeline.

Are there any post processing steps involved after TractSeg predicts the segmentation mask, before we calculate the Dice score?
The reason being, for HCP data, I followed all the preprocessing steps as given in this script. I changed the single shell CSD to multi-shell in the script. I got results for all test subjects - same as the ones used in the TractSeg paper. I use TractSeg version 2.8 to get the outputs for bundle segmentations.

However when I compute the Dice score - I get an overall average of 0.64. May be I am missing some pre/post processing steps?

Is there anything you can recommend so that I get the same Dice score of 0.84 as mentioned in the TractSeg paper?

Thanks a lot.