[Camino-users] Question regarding Monte Carlo Data Synthesis

[Ned Charles]

Hi everyone,

I have a question regarding Monte Carlo simulations using the datasynth module.  I am using datasynth to verify q-space type diffusion diffraction patterns for various synthetic meshes that we create as .ply files.  The simulation works great and creates the diffraction patterns, at least the first "resonance",? that match up to the literature for example: http://130.235.71.204/Publication/62_jmrA_1995_116_77.pdf

When looking at the created signal created under certain conditions, however, the actual returned signal from camino becomes negative in a couple of areas as the gradient strength increases.  To recreate the signal, the mesh substrate I am using is a 15 micron diameter sphere, with the file here: https://onedrive.live.com/redir?resid=C4C5DBA3D21E3ED6!149&authkey=!AE5S46TChx1pxLM&ithint=file%2c.ply

I am using a basic PGSE scheme with 101 different measurements at a small delta of 1 ms, a fixed big delta of 56 ms, TE of 66 ms, one gradient direction and an increasing gradient as in the scheme file here: https://onedrive.live.com/redir?resid=C4C5DBA3D21E3ED6!148&authkey=!APHn52CVBaXlHFo&ithint=file%2c.scheme

To replicate the experiment, I run datasynth on linux by:
datasynth -voxels 1 -walkers 10000 -tmax 1000 -p 0.0 -initial intra -voxelsizefrac 1 -substrate ply -plyfile <PLY FILE ABOVE> -schemefile <SCHEME FILE ABOVE> -outputfile Sphere.Bfloat

This should give a signal that has negative values in a couple of places.  The question I have is whether the returned signal from Camino in this instance is supposed to be a magnitude signal?  The actual signal appears to be built in Camino by taking the summation of the cosine value of the accumulated phase for each walker's trajectory, which I'm assuming is the "real" part of the signal.  I also browsed the code in AgnosticScan.java (version 2 DEC 13) and found that if an SNR is provided, a magnitude signal is created by taking the square root of the signal squared.  Upon testing this, it inverts the lobes on the signal and gives a "double dip", especially in a log plot vs. (q-value*sphere radius) as shown here: https://onedrive.live.com/redir?resid=C4C5DBA3D21E3ED6!150&authkey=!ANXxP6S9im65ACQ&v=3&ithint=photo%2c.png

This plot is for four experiments with different big DELTA values.  I also get the same results when saving the trajectories as a file and running different signals against that.  Please let me know if you aren't able to replicate the issue.  Just wondering if anyone can provide any guidance on the nature of the negative output signals and whether I am using the correct parameters to perform the experiment.  Any help would be appreciated.

Cheers,
Ned Charles
University of Sydney
ccha4217 at uni.sydney.edu.au

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.nitrc.org/pipermail/camino-users/attachments/20140314/e962e855/attachment.html>