Hi Derek,
No, it is the relative probabilities that are directly interpretable rather than the absolute number of streamlines.
Probtrackx estimates a spatial distribution of paths by drawing samples in a Monte Carlo fashion. You need a large number of samples to describe this distribution accurately (think of the definition of probability, it is the number of "successes" of an experiment, when it is repeated an "adequately large" number of times). Adequately large means enough samples to give converged probabilities.
The default number of 5000 samples is most of the times a good middle ground solution between convergence and execution time. If you convert the fdt_paths output to probabilities, you will see that in most regions the probabilities will be more or less the same, regardless of whether you use 5000 or 10000 samples. If you get a zero probability for 5000 samples at the lateral motor cortex, you should get an almost zero probability for 10000, at least in theory. In practice, you may get a very low probability instead reflecting imperfect convergence at 5000 samples.
Hope this helps,
Stam
On 18 Jun 2013, at 19:33, Derek Archer wrote:
> Hello,
>
> I was looking at my data, and I decided to change the "number of samples" option in probtrackx from 5000 (default) to 10000. Doing this, I was able to get lateral fibers.
>
> Would I increase the chance of getting false positives by increasing this number greater than 5000?
>
> Thanks,
> Derek
>
|