 |
 |
That was so fast! I really appreciate it. It leads me to 2 other questions: 1) Is there a time estimate on when the "fix" to bedpost will be ready (the one that will reduce the number of iterations from 5000 to 1250)? 2) I am baffled. How exactly did Tim recognize that the z-direction component was backwards in the data I sent? My slices are inferior to superior...is that an issue? How would I figure this out myself? Thankyou SO much as always....what incredible software and incredible support! Dianne On 8/21/07, Matt Glasser <[log in to unmask]> wrote: > > Yes to all three. J > > > Peace, > > > > Matt. > > > ------------------------------ > > *From:* FSL - FMRIB's Software Library [mailto:[log in to unmask]] *On > Behalf Of *Dianne Patterson > *Sent:* Tuesday, August 21, 2007 12:58 PM > *To:* [log in to unmask] > *Subject:* Re: [FSL] new bedpost questions > > > > Wow, thankyou Tim, for the evaluation! > If I may ask for clarification, however: > 1) The z-vector is the third row in my bvecs (correct?) > 2) So, you are suggesting that I switch the sign of each value there, like > so: > If these are my values now: -0.840197 0.64528 -0.177358 > Then I switch to: 0.840197 -0.64528 0.177358 > Is that correct? > 3) Should I also rerun dtifit and bedpost after repairing bvecs? > > -Dianne > > On 8/21/07, *Tim Behrens* <[log in to unmask]> wrote: > > Hi Dianne - > > I am glad it looks better. > > However, I was looking at your data and I noticed that something wasn't > quite right. > > > > I think the reason you have been successful in tracking the dorsal portion > of the SLF, is that this pathway lies predominantly in the XY-plane. I think > if you tried any pathway that had a reasonable Z component, you would find > that your results looked very odd indeed (for example the SLF III does not > descend into the temporal lobe as it often does). > > > > I think the reason for this is that you have defined your Z-component of > your bvec in the opposite way from the way that FSL defines it. > > > > I think if you flipped this around (just take the negative of all the z > directions in bvecs) that you would find that your results were even more > impressive. > > > > I think this is also the reason that the second (and first..) directions > do not quite line up as they should. > > > > Cheers > > > > T > > > > On 20 Aug 2007, at 20:41, Dianne Patterson wrote: > > > > The upload *DP.tar.gz* is 155213. > It contains a single and double fiber Feeds run and my bedpostx > directory. > It has a couple of layers of tar files...whose names should be fairly > transparent. > > Additionally, I sent *SLF_Prob.tar.gz* which contains the 2 probability > trackings for the superior longitudinal fasciculus which I described in an > email the the listserve a few minutes ago: > > Dear group, > > Just to share what I consider a great success: > > Here we compare the old fsl 3.3 analysis with one fiber path (called > onefdt_paths.nii.gz) > to the new fiber track (fdt_paths.nii.gz) created with the same slf28a > mask and the default > of 2 fiber paths. > > dpat% fslstats onefdt_paths.nii.gz -V -M > 2423 42.637.746094 1225.950887 > > dpat% fslstats fdt_paths.nii.gz -V -M > 8593 151.211.781250 461.366694 > > Note the new area resulting from probtrackx is triple the size... > It covers all the old territory that the single fiber tracking identified, > AND it identifies more real white matter as seen in the FA. > It identifies longer paths and more branches of the slf > than the original single fiber analysis. > > The data is 25 directions at B=1000 and 3 B0 images > gathered on a GE 3.0 T scanner with parallel imaging and an 8 channel > coil. > The matrix was 128x128 with roughly isotropic voxels of 2.6x2.6x2.6 > Rough calculations suggest a SNR of about 20. > > I hope I haven't left out any crucial details!! > > -Dianne > > Dianne > ================= > > On 8/20/07, *Tim Behrens* < [log in to unmask]> wrote: > > Hi Dianne - If you would like to upload the bedpostX directory of your > data - we can have a quick look. > > > > While you're at it, if you upload the bedpostX directory from the course > data, it will save me running it before the FSL course :) > > > > www.fmrib.ox.ac.uk/cgi-bin/upload.cgi > > > > Cheers > > > > Tim > > > > > > On 17 Aug 2007, at 20:42, Dianne Patterson wrote: > > > > > > On 8/17/07, *Tim Behrens* < [log in to unmask] > wrote: > > > > For 30 directions, It looks good in simulation for voxels where the > fibres are very separated, but simulation is simulation! > > > > Johannes Klein, in our lab, wanted to use it for clinical populations, and > seemed happy with its performance with 30 direction data (do you remember > anything else about the protocol, J?). > > > > Good to know, thankyou...my output LOOKS pretty...is there any other way > to evaluate it? > > > > I suppose you could compare the 2nd directions with a 60 direction > dataset (e.g. the FSL course data). > > > > I think if the 2nd fibre orientations line up to make pathways, you should > think of this as impressive - there is no reason why they would if you were > just looking at noise. > > > Okay I ran bedpostX on the feeds data and tried to align that slice with > my data to compare...my subjective impression is that the 60 direction data > has more coherent paths, but I have some in my data...though the paths tend > to be shorter in my data and perhaps not quite so consistently pointing in > the same direction as their partners...(of course you all are welcome to > pictures or the data files if you'd like, but I imagine you are a tad busy > with changes that follow on a new release...so I'll just keep plugging away > with probability tracking and see if I can compare results. > -- > Dianne Patterson, Ph.D. > [log in to unmask] > ERP Lab > University of Arizona > 621-3256 (Office) > > > > > > > -- > Dianne Patterson, Ph.D. > [log in to unmask] > ERP Lab > University of Arizona > 621-3256 (Office) > > > > > > > -- > Dianne Patterson, Ph.D. > [log in to unmask] > ERP Lab > University of Arizona > 621-3256 (Office) > -- Dianne Patterson, Ph.D. [log in to unmask] ERP Lab University of Arizona 621-3256 (Office)