Hi Darryn,

I was reading the forum and noticed that although I replied to your question minutes after you posted it back on February 24th, my reply was never posted. Theoretically, I should be able to reply to a question through my e-mail program if I put my answer between some markers. Although that e-mail was sent, my answer never made it to the forum.... I apologize for that. Here is what I wrote back then. I hope it is still helpful.

Hi Darryn,

The threshold is totally up to you. Since these bundles were generated with the two-ROIs approach, you don't have to worry much about stray fibers that you will need to reject by raising that threshold. Raising the threshold will mainly make the bundles thinner, and if you use them as ROIs, you will simply avoid some partial volume effects at the edges of the bundles. Just try different thresholds, then bring these ROIs to the native space of each of your subjects and see if you need to threshold more aggressively or not.
As to your second question, you are correct. You typically focus in a certain bundle if you have a hypothesis about that bundle. If you would like to investigate the whole brain, you would not select ROIs defined by a collection of bundles. You would probably do a voxel-wise tbss type of analysis.
Now, since you asked about thresholding the bundle files and I imagine you may be worried about partial volume effects in ROI analyses using atlas-based segmentation vs. tbss analyses where you project information from the central portion of white matter tracts, I would like to point out the option of skeletonized atlas-based segmentation (see http://www.ncbi.nlm.nih.gov/pubmed/24925...). The "major bundles" files of the IIT atlas are also available in skeletonized format. This allows you to project the information from your subjects to the skeleton and then extract information from the portion of the skeleton corresponding to the large bundle of interest. This approach has advantages over the conventional atlas-based segmentation approach in that it reduces partial volume effects.
I hope this helps.
Konstantinos