[Neuroimaging] [Dipy] Warnings when running free water model

[Ariel Rokem]

Hi Arnold,

Thanks for your email.

On Thu, Dec 28, 2017 at 8:11 PM, Arnold Evia <ajevia at gmail.com> wrote:

> Hello,
>
> I am applying the free water dti model to my data, and receive the
> following warnings:
>
> /Users/arnoldevia/anaconda3/lib/python3.6/site-packages/scipy/optimize/minpack.py:427:
> RuntimeWarning: Number of calls to function has reached maxfev = 1800.
>   warnings.warn(errors[info][0], RuntimeWarning)
>

This warning means that when fitting the model to the data, the software
encountered a voxel where the data looks very different from what is
expected for this model, and the optimization required to find the model
parameters terminated before it could find good values for the parameters,
because it is set to stop after trying that many times.

This is not usually something to worry about -- unless it happens with
every single one of your voxels, and will arise in boundaries of the brain
with other tissue (e.g., skull or dura), where the signal looks nothing
like what you would expect for the model. I would look for extreme outliers
in the model parameter values and examine where they occur (like you might
look for negative eigenvalues, if you were to fit the DTI model).

/Users/arnoldevia/personal_programs/dipy/dipy/reconst/fwdti.py:454:
> RuntimeWarning: overflow encountered in exp
>   y = (1-f) * np.exp(np.dot(design_matrix, tensor[:7])) + \
>
>
I believe that this one occurs under similar conditions -- when the model
parameters become very extreme, and the exponent goes to very high values,
far beyond what you might expect in the physiological case. Again, I would
look for extreme model parameter values in the results, and try to see
whether they occur at tissue boundaries, or in non-brain/non-CSF tissue.

Hope that helps!

Ariel

The code I am using is essentially the same as the code used in the free
> water model documentation.
> What do these warnings mean, and what should I do to address them?
> Please let me know if additional information is needed.
>
> Thanks in advance!
> -Arnold Evia
>
> _______________________________________________
> Neuroimaging mailing list
> Neuroimaging at python.org
> https://mail.python.org/mailman/listinfo/neuroimaging
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/neuroimaging/attachments/20171229/c000166f/attachment.html>