[Neuroimaging] Regarding PIESNO Paper by Koay

Samuel St-Jean stjeansam at gmail.com
Thu Apr 28 07:36:06 EDT 2016

As a first point, piesno actually works only (based on the math) for rician
or non central chi distributed noise (which is what is commonly found in
magnitude images in MRI) and correct for that case to get back an
estimation of the standard deviation based on a single gaussian
distribution. You could have a look at [1] to get a feeling of the various
noise distribution encountered in most cases, as the gaussian case would
only apply if you also collect phase data I'd say (or do fancy
pre-processing scanner side, which is not commonly done by everyone).

Anyway, to go more toward your question, you can have at [2] which gives
method to estimate the number of coils of the acquisition. Although these
require knowledge in the type of acquisition you have, it does lay some
groundwork for ideas on how to estimate the noise from any acquisition. I
also tried to implement the method of [3] (probably have some code lying
around if I look hard enough if you would want it) but found the estimation
to not be inline with my simulations (could be also my quick
implementation, I had nothing to compare against).

Talking about implementation, if you want to validate/add features to it,
there is alos another piesno implementation over here :

This was validated against the original closed source code, but as you can
see I used the precomputed tables from the article. If you also have the
original equations implemented instead of the precomputed value it would be
a great addition (since it requires 1D optimisation, I though it would be
too slow to compute it each time, so you might also want to use these
tables instead in that case.)

Anyway, if you need more help feel free to ask back.


[1] Dietrich et al. 2008 Measurement of signal-to-noise ratios in MR
images: Influence of multichannel coils, parallel imaging, and
reconstruction filters
[2] Aja-Fernández, S., Vegas-Sánchez-Ferrero, G., Tristán-Vega, A., 2014.
Noise estimation in parallel MRI: GRAPPA and SENSE. Magnetic resonance
imaging 32, 281–90.
[3] Veraart et al. 2013 Comprehensive framework for accurate diffusion MRI
parameter estimation

2016-04-28 9:44 GMT+02:00 Vivek Joshi <vivekjoshi1894 at gmail.com>:

> Respected Sir,
>                       My name is Vivek Joshi and i am doing a project on
> PIESNO by Koay for my final year project. We implemented the PIESNO code in
> python. But our Professor is asking us to identify the noise (whether
> rician or gaussian) present in mri images. So it would help me if you have
> any idea on how to identify noise. Please think about it.
> _______________________________________________
> 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/20160428/967f4b71/attachment.html>

More information about the Neuroimaging mailing list