I've created a couple of examples, I guess it will aid diagnosing:

-> The first schedule is here: https://gist.github.com/oesteban/8c93a6553125d17c8990

Basically, I remove the standard 8mm scale (and I think that for very small translations/rotations even the 4mm could be removed). I also increased the smoothing at each level.

-> The second schedule is here: https://gist.github.com/oesteban/6e3d8777d4e1ea728f17

In this case, it is pretty much the same as before, but now I add the option 'setoption paramsubset 1  0 0 0 0 0 0 1 1 1 1 1 1' in order to only optimize scalings and skews.

My problem is that this language is a bit difficult to understand. Basically I'm testing by chance.

I think that, when one inserts the optimise command, the 'amount' of transformation is controlled by the last part (something of this kind: 'rel 4') but I'm not sure of the implications of this parameter.

e.g. optimise 7 UA:1-4  1.0   0.0   0.0   0.0   0.0   0.0   0.0  rel 4

I'm also unsure about the first schedule, around the end, when it tries again the identity transform. I'm not sure this is taken into account (these copy U UA commands are quite obscure to me, sorry).

Thank you so much

Best wishes,

Oscar

______________________________________

Oscar Esteban

PhD Student / Researcher

Biomedical Image Technologies (BIT), UPM
ETSI Telecomunicación Lab. C203, Av. Complutense s/n - E-28040 Madrid (Spain)
+34 915 495 700 ext.4234

Signal Processing Laboratory (LTS5), EPFL-STI-IEL-LTS5
ELD 224 (Bâtiment ELD), Station 11, CH-1015 Lausanne, Switzerland

On 4 September 2014 18:16, Oscar Esteban <[log in to unmask]> wrote:

Sorry, I wrote schema where I should have said 'schedule'

______________________________________

Oscar Esteban

PhD Student / Researcher

Biomedical Image Technologies (BIT), UPM
ETSI Telecomunicación Lab. C203, Av. Complutense s/n - E-28040 Madrid (Spain)
+34 915 495 700 ext.4234

Signal Processing Laboratory (LTS5), EPFL-STI-IEL-LTS5
ELD 224 (Bâtiment ELD), Station 11, CH-1015 Lausanne, Switzerland

On 4 September 2014 16:59, Oscar Esteban <[log in to unmask]> wrote:

Dear all,

I'd like to write a couple of schemas in order to perform registration with FLIRT:

-> One for motion correction in dMRI datasets (6 dof, very small rotations/translations). I see that search angles can be limited, what about translations?.

-> One for eddy currents correction (12 dof, nosearch, no translations): here, my doubt is if I can forbid translations during optimisation.

My first try will be to re-use one log-file from flirt, executed as in the old eddy_correct script and then set there my options. I think I will be able to easily forbid translations and rotations, looking at the other files. However, I don't know how to limit the amount of translation.

Best

Oscar Esteban

______________________________________

Oscar Esteban

PhD Student / Researcher

Biomedical Image Technologies (BIT), UPM
ETSI Telecomunicación Lab. C203, Av. Complutense s/n - E-28040 Madrid (Spain)
+34 915 495 700 ext.4234

Signal Processing Laboratory (LTS5), EPFL-STI-IEL-LTS5
ELD 224 (Bâtiment ELD), Station 11, CH-1015 Lausanne, Switzerland