 |
 |
Dear James, > After running tbss_2_reg I get several log files in /FA/tbss_logs saying, e.g.: > > "Warning, Jacobian not within prescribed range. Prescription is 0.01 -- 100 and obtained range is -8.68585 -- 24.7955" > etc. this is dramatically outside the prescribed range and something else must be going on. My comment below refers to cases where there is some small deviation between the prescribed and obtained lower bound. The first thing you should do is to find out which scan this message refers to and look at the output of the registration. My prediction is that it will be warped up like a discarded crisp bag. Then you need to look at the original file for that scan and see if you can work out why things go wrong. Chances are there will be some problems/artefacts in there. If you still can't work out what is the problem with the scan, please get back to us. Jesper > > Some logs have one such message, others have several (e.g. 4). > > I see in a message from October 2008 that Jesper Andersson spoke of altering the lambda in fnirt (quoted below). So presumably I need to do some corrections and re-run this step? > > 1. Do I need to change 'lambda' somehow for this tbss_2_reg set of error messages, and if so, how would I do that? > Or is there some other way to deal with this issue? > > 2. Depending on the answer to (1), do I need to set something to account for the specific range of values that appear in the error logs? > > Cheers > James > > Jesper Andersson: >> Warning, Jacobian not within prescribed range. Prescription is 0.01 -- 100 >> and >> obtained range is -0.00444645 -- 3.01544 > "After each iteration fnirt attempts to make sure that the Jacobians > (stretches and compressions) are within a pre-determined range. You have > used the default range, which basically just assures that the field is > invertible. When doing that fnirt runs a fixed number of iterations, and > if it hasn't succeeded after those iterations it prints the error message > above. > > For the next release I'll change that so that one can set the number of > iterations (to make sure it "always" succeeds). For the time being I would > recommend that you slightly increase the amount of regularisation that you > are using. I.e. in your config file you increase the values for lambda. > > If you are using the T1_2_MNI152_2mm file you might e.g. change lamda to > > --lambda=400,200,150,75,60,45 > > and I would expect the problem to go away" >