Tcksift warnings

tckgen …/WM_FOD.mif 10M.tck -seed_image …/…/b0_brain_mask.nii.gz -select 10M
tckgen: [100%] uncompressing image “…/…/b0_brain_mask.nii.gz”
,10 -exclusive 1M.tck path tck_roi/tck
tckgen: [100%] 10967223 seeds, 10729185 streamlines, 10000000 selected
[liuyc@head01 tournier]$ tcksift 10M.tck …/WM_FOD.mif 1M.tck -term_number 1M
tcksift: [100%] Creating homogeneous processing mask
tcksift: [100%] segmenting FODs
tcksift: [100%] mapping tracks to image
tcksift: Iteration Removed Remaining Cost fn
tcksift: [ . ] 2344 18 2427589 5.28%
tcksift: [WARNING] filtering has reached quantisation error but desired termination criterion has not been met;
tcksift: [WARNING] disabling cost function quantisation check
tcksift: [done] 13559 0 1245815 4.01%
tcksift: [WARNING] algorithm terminated before any user-specified termination criterion was met
tcksift: [100%] Writing filtered tracks output file

Here are warnings I have received when performing tcksift, while the number of sifted tracks is still below 1M, are those right tracks?

Another question is that, when I calculate the correlation of two subjects’ ROI-ROI (AAL2) tracks number, the pearson’s correlation is very high (1-2,3-4,5-6 are same person, the number are their correlation ), it seems unreasonable to me, what could be possible reasons for this high correlation?
image

Here are warnings I have received when performing tcksift , while the number of sifted tracks is still below 1M, are those right tracks?

Seeing the phrase “right tracks” makes me uneasy :upside_down_face:

The terminal output is faithful to what’s going on, if expressed in a more general way. The SIFT algorithm is literally unable to find a single streamline that it predicts will lead to a decrease in the cost function. So instead of removing streamlines and increasing the cost function, it instead terminates with a greater number of streamlines remaining than what you requested. The “quantisation error” one is a little more cryptic, but the criterion is described in the relevant manuscript.

Another question is that, when I calculate the correlation of two subjects’ ROI-ROI (AAL2) tracks number, the pearson’s correlation is very high (1-2,3-4,5-6 are same person, the number are their correlation ), it seems unreasonable to me, what could be possible reasons for this high correlation?

Is this literally the Pearson correlation coefficient calculated across all edges in the full AAL2 atlas? If so, this is the result of calculating a Pearson correlation coefficient on data that vary by many orders of magnitude; it is simply not designed to handle such data. If anything, performing a Pearson correlation on the logarithms of those values would make more sense, since the noise in structural connectomes is approximately multiplicative.