In principle, increasing the number of iterations etc - as well as the
settings to improve the accuracy of the multigrid should help,
although I suspect that this will not be by much. The greatest
improvements are likely to be through figuring out the optimal
regularisation. Ideally, this should model our prior knowledge of
anatomical variability.
One other SPM8 approach for increasing accuracy (slightly) would be to
use the more mathematically correct Geodesic Shooting toolbox.
However, this was only released in a late update of SPM8 and it is
still a bit lacking in the sort of functionality that would make it
easy for a non-expert to use.
Best regards,
-John
On 12 March 2012 11:32, Hikaru Takeuchi <[log in to unmask]> wrote:
> Dear Prof. John Ashburner and SPM users,
>
> Thank you very much for the information.
> I am now trying to sophisticate the process and would like to change the parameters in DARTEL after all.
> In the case of this process, actually there were some instances that the alignment qualities were desired to be improved a little bit in some areas for what ever the reasons.
> So, I would like to change some parameters in the DARTEL procedures to improve the accuracy at the cost of time.
> In that case, looking at the explanation of the DARTEL and from what I understand, it seems changing the parameters as follows increases the accuracy, am I correct? Or it is not so simple?
> Outer Iterations->Outer iteration->Inter Iterations - increase as much as possible (in the practical amount of computation time) in each Outer Iteration (this is not clarified in the help and I am not sure)
> Outer Iterations->Outer iteration->Time Steps - increase as much as possible in each Outer Iteration
> Optimization Settings->LM Regularization - increase as much as possible until 1
> Optimization Settings->Cycles - increase as much as possible
> Optimization Settings->Iterations - increase as much as possible (this is not clarified in the help and I am not sure)
> Some of these may go without saying and I am sorry if the questions are silly. But I do not understand what's going on in the processes, and am happy if you would tell me if I am wrong.
> Best regards,
>
> Hikaru Takeuchi
>
>
>
|