6.1. Selection of Beta

An appropriate choice of the ADMM/LADMM parameter (defined as beta in this software) is important for fast convergence. We have provided functionality to help users compare the convergence speed for different beta values over small portions of the image, so that an appropriate choice can be made quickly and efficiently, prior to estimating the full image. To use this tool, the user needs to create a spatial mask in spatmaskfile format defining a small image patch, and then run the following command.

plot_beta_sweep.sh -imgfile  imgfile_name -spect_infofile spect_infofile_name -configfile configfile_name -spatmaskfile spatmaskfile_name -outprefix outprefix_name

plot_beta_sweep('imgfile','____', 'spect_infofile', '____','configfile','____','spatmaskfile','____','outprefix','____',)

Note

  • This command chooses beta values based on the solver.beta value present in the configfile. Then, the beta values chosen are \([0.01\;0.1\; 1\; 10\; 100] \times\) solver.beta.

This command saves out a figure in .png format showing the convergence speed and a zoomed out version of it. Below is an example output.

Workflow overview

The user can also input the following optional argument with user defined beta values and file types of the stored figure.

plot_beta_sweep.sh -imgfile  imgfile_name -betafile betafile_name -spect_infofile spect_infofile_name -configfile configfile_name -spatmaskfile spatmaskfile_name -outprefix outprefix_name -file_types file_types_names

plot_beta_sweep('imgfile','____','betafile','____', 'spect_infofile', '____','configfile','____','spatmaskfile','____','outprefix','____','file_types', '____')

6.1.1. Required Inputs

  • imgfile:

    Filename of the .mat file storing the MR data related information. Details of the file format can be found here. The user must create this file before proceeding with beta parameter calculation.

  • spect_infofile:

    Filename of the .mat file containing spectrum-related information. The user needs to create this file based on user selected spectral points and should follow the fileformat mentioned here. This tool can handle multiple spectral dimension (including 1D spectra) as long as this file is stored in the given format.

  • configfile:

    Filename of the .ini file storing spectrum estimation algorithm parameters. The spectrum estimation can be done using three algorithms -LADMM, ADMM, NNLS. The file format is described here.

  • spatmaskfile:

    Filename of the .mat file storing the binary spatial mask for the MR input data. This specifies voxel locations where spectrum estimation would be done for multiple beta values. The spatial dimension of the variable in this file must match with the spatial dimension of the input data. See here for file format.

Note

  • As the spectrum estimation algorithm with multiple beta values may take long time to run if whole MR volume is chosen, its advised to create the spatial mask as multiple \(3\times3\) patches of voxels per slice. Also its best to choose the patched in location where the most partial voluming occurs. The user can use our GUI based patch selection tool.

  • outprefix:

    A string input for the filename where the output of the beta calculation tool will be stored. By default it saves a .png file. It stores two images in one image file- first image show the cost function vs iteration number for multiple beta values. The number of iteration depends on when the solver.tol or solver.numiter is reached. The seconf image shows a zoomed in version of the same image, but the x axis is chosen to show values from the halfway to the final number of iterations and y axis is chosen \([0.99*min_c\; 1.05*min_c]\) , where \(min_c\) is the minimum of all the cost functions.

6.1.2. Optional Inputs

  • betafile:

    This is a .txt file containing the beta values. This file should be written in the format mentioned below and the variable name should always be beta_vals in the .txt file.

    beta_vals= [0.05 0.5 1 5];

  • file_types:

    This is an optional input where user can spacify the file types of the final stored image. The input must be a char, string, cellstr or string array. Available file_types: fig, m, eps, epsc, ai, pdf, bmp, jpg, jpeg, png, tif, tiff. Default: png. Invalid entries are ignored; if all are invalid, falls back to png