Combine two successive MPM datasets

The running of this module requires the outputs of Create Maps module which are generated with a specific local defaults file. In the second step, these intermediate outputs are used in this module to combine two successive MPM datasets. We clarify the details below.

Generate error maps:

The error maps obtained here give the voxelwise error of the quantitative maps for quality assurance. Thereby, an evaluation of local data variations and artifacts are obtained without a need for additional data (Mohammadi et al. 2022).

In the Create Maps module, by using the following hmri local defaults file

hmri-local-defaults-errormaps

which turns on the parameter hmri_def.errormaps on this line

hmri_def.errormaps = true;

one can, in particular, obtain the following additional maps in the Results/Supplementary directory

anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_PD_param_error.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_R1_param_error.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_MTsat_param_error.nii

These error maps now can be used in the Combine two successive MPM datasets module, in order to obtain a robust combination of two acquisitions.

Combine two successive MPM datasets:

This submodule uses the above error maps as weights to robustly combine quantitative maps from two successive acquisitions, resulting in an arithmetic mean, a robust combination and an error map (for the robust combination) for each pair of the provided MPM maps (Mtsat, PD, R1). The resulting maps show better performance in variability compared to the single repeat or averaged counter-parts (Mohammadi et al. 2022).

On the Matlab batch, the inputs are:

• Output directory choice
• MTsat, PD and R1 maps from first run
• MTsat, PD and R1 maps from second run
• Weight images (error maps) for first run
• Weight images (error maps) for second run
• Reference image (optional)
• Brain mask (optional)

If no reference image is entered, by default the map from the first run will be used as a reference to which all data will be resampled.

Upon successful completion of the module, located in the (above) chosen output directory are the following images:

anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_MTsat_AM.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_MTsat_param_error_RO_errormap.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_MTsat_RO.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_PD_AM.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_PD_param_error_RO_errormap.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_PD_RO.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_R1_AM.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_R1_param_error_RO_errormap.nii
anon_s2018-02-28_18-26-185345-00001-00224-1_RFSC_R1_RO.nii

where the naming conventions _AM, _RO and _RO_errormap respectively denote the arithmetic mean, robust combination and robust combination error maps of the two successive acquisitions of the corresponding MPM maps (MTsat, PD, R1).

Notes on parameters in the hmri-local-defaults-errormaps:

The following parameter

hmri_def.wcombparams.kt         = 10; 

might need to be adjusted based on the dataset and the protocol. In a nutshell, the percentage parameter kt = hmri_def.wcombparams.kt/100 provides the steepness argument in the calculation of the local Fermi function on this line:

function f = local_fermi(x,kt)

Examples of how to choose this parameter can be found in (Mohammadi et al. 2022, Supplementary Materials: S1: Efficiency of robust combination and the Fermi function).

References

• Mohammadi, Siawoosh, et al. "Error quantification in multi-parameter mapping facilitates robust estimation and enhanced group level sensitivity." NeuroImage 262 (2022): 119529.
• Original pull request: https://github.com/hMRI-group/hMRI-toolbox/pull/49