Mark,
In the context of "slice by slice" registration, this will give you a
blank output slice (presumably because of this tiny z translation).
flirt -in $frame$slice -ref $b0mean$slice -applyxfm -init $matAtoD -out
$frame"_xenc"$slice
Regards
Gordon
On 04/23/2010 03:45 AM, Mark Jenkinson wrote:
> Hi,
>
> Yes, there are some rounding errors in the matrix libraries that we use
> which generate tiny numbers like this. Given that it represents a
> translation
> of 0.61 nanometres, I normally find that this doesn't make any
> difference to
> anything - even when concatenated with other transformations with similar
> rounding errors. Do you have an example of where this is actually
> important?
>
> All the best,
> Mark
>
>
> On 23 Apr 2010, at 08:38, Junqian Gordon Xu wrote:
>
>> flirt -in $frame$slice -ref $frame"1"$slice -out $frame"_xenc"$slice
>> -omat $tempmat -2D
>>
>> $tempmat:
>>
>> 0.989301 -0.145891 0 3.07181
>> 0.145891 0.9893 0 -5.26495
>> 0 0 1 6.10647e-07
>> 0 0 0 1
>>
>> When running a slice by slice flirt -2D registration, the output mat
>> sometimes (seems randomly) contains a small z translation (e.g.
>> 6.10647e-07, see above). The output $frame"_xenc"$slice is fine though.
>>
>> This z translation causes trouble when one tries to concatenate
>> transformations later on. I currently use the following line to reset
>> the z-trans value back to 0.
>>
>> awk 'NR==3 {$4=0}1' $tempmat > $mat
>>
>> I don't think it's of a high priority to be investigated or fixed, but
>> thought it's good to point it out for people going down the same route.
>>
>> Regards
>> Gordon
>>
>
|