 |
 |
Thank you for the succinct information Mark! I really appreciate it. Abhinay On Mon, Sep 11, 2017 at 6:12 AM, Mark Jenkinson < [log in to unmask]> wrote: > Hi, > > > Is there a link that helps to identify the default parameters in FLIRT? > > Not really - but I'll describe it here and then add it to our wiki. > > > I am trying to find out what does the following parameters do in the > FLIRT registrations? > > > > -searchrx <min_angle> <max_angle> (angles in degrees: default is -90 > 90) > > -searchry <min_angle> <max_angle> (angles in degrees: default is -90 > 90) > > -searchrz <min_angle> <max_angle> (angles in degrees: default is -90 > 90) > > These parameters determine the range of starting points for the > multi-start search phase of our optimisation algorithm. They do not > constrain the final angles found by the optimisation, just the starting > points. However, if the images start with orientations that are relatively > well aligned already then it is useful to restrict this range to avoid > initial guesses that are a long way off the truth. Our defaults are quite > wide and work for most cases, with minimal assumptions, but if you don't > need such large ranges then you can increase the robustness by reducing it > (if appropriate). > > > -coarsesearch <delta_angle> (angle in degrees: default is 60) > > -finesearch <delta_angle> (angle in degrees: default is 18) > > These control the search phases, in terms of how many multi-start points > to choose, and how they are spaced. As indicated by the names, there is an > initial coarse search phase and then a fine search phase later on. I have > very rarely (if ever) had to change these, but it is possible that reducing > these might help (with a trade off with execution time). > > > The reason I ask is when I try to use FLIRT 12 parameter affine to > register native space PET TAU images to a template image created from the > same tracers, it works on 95% of the cases. However, when I use SPM 12 (old > normalize routine) the success rate is lower i.e. 80%. I was able to fix > the 7% of the 20% failed cases from SPM12 by fixing their origin. But there > are still 13% of cases failed registration because of high scalp uptake, > hot calivus or hot salivary gland uptake. So I may have to deskull. > > You could try to apply a cost function weighting to deweight areas where > you are seeing problems (calivus or salivary gland). This would probably > involve a registration of the MRI to the MNI template and then the inverse > transformation of a cost function weighting mask (with zeros in these areas > to deweight). > > It might be worth doing brain extraction (deskulling) but only if you can > do it to both images. > > I would also recommend fslreorient2std and cropping (via robustfov or > manually with fslroi) to see if that fixed things (assuming that your PET > images have the appropriate orientation information - if not then this > might be the cause of some problems). > > > Where as FSL FLIRT performs better on 95% of the cases, which tells me > FSL FLIRT is figing the origin issues and seems like more robust, which > could be because of the above parameters. > > I think that some combination of the above might help. > > > If anyone could help me understand the FLIRT parameters better that will > be a huge help. > > I hope this is helpful. > All the best, > Mark > > > > > Thanks. > > -- > > Abhinay D. Joshi > > Imaging Research Advisor > > Avid Radio-pharmaceuticals Inc. > > Philadelphia PA. > > > -- Abhinay D. Joshi Wayne, PA Cellular:-817-995-3962