jist:Noise Field Estimation

Robust Estimation of Spatially Variable Noise Fields

Here, we provide a layout file for duplicating the functionality presented for a repeated acquisition of a set of images with differing contrast, but with the same fundamental noise properties.
All tools used in these files have been released in open sourced on NITRC and are included with the default binary JIST distribution.
You may replicate this functionality exactly by placing modules from the JIST toolbox in your layout.

B. Landman, P. Bazin, S. Smith, J. Prince, Robust estimation of spatially variable noise fields, Magnetic Resonance in Medicine, Volume 62, Issue 2, Pages 500 - 509

B. Landman, P. Bazin, J. Prince, Estimation and application of spatially variable noise fields in diffusion tensor imaging, Magnetic Resonance Imaging, Volume 27, Issue 6, Pages 741-751

Requires JIST Beta 1.2 or newer
Sample Data Available Here: http://www.nitrc.org/frs/download.php/1009/020729.tar.gz Data use license: 'These data are freely available for research purposes. For publication, please identify the source of the data as "Data were obtained as courtesy of Dr. Mori (www.mristudio.org) in Johns Hopkins University."'
Run CATNAP on the data. Use "tar -xvzf" to decompress this dataset.
Download the Noise Field Estimation layout file: http://www.nitrc.org/frs/download.php/1050/NFestimation-2009August27.layout
You may save layout file and/or the sample data in any location that is accessible to MIPAV.

Select the files that correspond to the first repetition it the top box and the second repetition in the lower box.

Note: You may select any combination of 3D or 4D volumes. The application concatenates the volumes along the 4th dimension in the order specified in the box. The repeated pairs of volumes must be in the same relative location in each box.
In the number of signal averages box, indicate how many signal averages were taken for each volume.
If all volumes have the same number of averages, you may enter just a single number. Otherwise, the number of listed entries must match the number of volumes.
Select the regularization module. You may adjust the polynomial degree of the regularization field in the selection box.

Select Project->Layout Preferences (1) and select an output directory (2). Optionally, adjust the maximum amount of memory or maximum number of simultaneous cores that will be run.

Once the process manager has finished, you may open the local noise field estimate, local confidence weighting, and regularized field.

B0 image slice

Local noise field estimate

Relative Noise Field Confidence

Regularized noise field

Important: The resulting noise field estimates are in the units of the original image space. No conversion is done to calculate SNR.
That's it.

Potential issues:

The b-value files should end in .b (on the disk and in the list file)
The b-values should be in a single column and named <filename>.b
The gradient table should be in 3 columns and named <filename>.grad
Please be sure that the full paths listed in the text files is actually correct.
On some systems, support for compressed image files can cause trouble. If you are having trouble, try decompressing the image files and running the module again. If this is the root of a problem with your data, please let us know (on the bug-report forum --- see "Tracker" at left) so that we can resolve it.