complexSVDShrinkageDWI

Tools for complex DWI denoising using SVD shrinkage

This repository provides tools to implement the methods and reproduce the experiments included in the manuscript ''Complex diffusion-weighted image estimation via matrix recovery under general noise models'', L Cordero-Grande, D Christiaens, J Hutter, AN Price, and JV Hajnal, arXiv:1812.05954.

The code has been developed in MATLAB and has the following structure:

./

contains the scripts for running the experiments included respectively in Figs. 4, 9, 6 and 5 of the manuscript: complexSVDShrinkageDWI_Exp[1-4].m, generate part of the graphical materials included in the Figs.: plot_Exp[1-4].m, call the patch-based traversal: patchSVShrinkage.m and perform SVD shrinkage: SVShrinkage.m.

./Build

contains scripts that replace, extend or adapt some MATLAB built-in functions: diagm.m, dynInd.m, eigm.m, emtimes.m, ind2subV.m, indDim.m, matfun.m, multDimMax.m, multDimSum.m, numDims.m, parUnaFun.m, resPop.m, resSub.m, sub2indV.m, svdm.m.

./ESDEstimation

contains scripts for empirical sample distribution computation: ESDMixAndMix.m, ESDSimulated.m, ESDSpectrode.m.

./ESDEstimation/MixAndMixCode

contains scripts that implement the MidAndMix ESD computation method in the manuscript ''Numerical techniques for the computation of sample spectral distributions of population mixtures'', L Cordero-Grande, arXiv:1812.05575: andersonMixing.m, arraySupport.m, fillGridPoints.m, gridSubdivide.m, interp1GPU.m, nonUniformGridAddPoints.m, pinvmDamped.m, startingGrid.m.

./ESDEstimation/SpectrodeCode

contains scripts that implement the Spectrode ESD computation method in the manuscript ''Efficient computation of limit spectra of sample covariance matrices'', E Dobriban, Rand. Matr. Th. Appl., 2015, 4(4):1550019:1-36 from https://github.com/dobriban/EigenEdge/

./Figures

contains scripts for generating the graphical materials: genericFigInformation.m.

./Lib

contains external MATLAB tools.

./Lib/Colormaps

contains tools to generate colormaps from https://uk.mathworks.com/matlabcentral/fileexchange/51986-perceptually-uniform-colormaps.

./Lib/line_fewer_markers_v4

contains tools for plotting from https://uk.mathworks.com/matlabcentral/fileexchange/42560-line_fewer_markers.

./Lib/NIfTI_20140122

contains tools for tools for NIfTI and ANALYZE formats from https://uk.mathworks.com/matlabcentral/fileexchange/8797-tools-for-nifti-and-analyze-image

./Matrices

contains scripts for operations with matrices: mat2bldiag.m.

./Methods

contains scripts that implement generic methods for reconstruction: build1DFTM.m, buildStandardDFTM.m, fctGPU.m, fftGPU.m, generateGrid.m, ifctGPU.m, ifftGPU.m, mirroring.m, plugNoise.m, resampling.m, ridgeDetection.m, spaNeigh.m.

./Shrinkage

contains scripts for SVD shrinkage: frobenius.m, generalShrinkage.m, hard.m, operator.m, percMarcenkoPastur.m, stieltjes.m, stieltjesSimulated.m, veraart.m.

NOTE 1: Exemplary data is provided in the datasets recFig0[4,6a,6b,9].mat, dwi[w,m,x,y].nii, dw_scheme_b10000.[bval,bvec]. For runs without changing the paths, they should be placed in a folder

../complexSVDShrinkageDWIData

Data generated when running the scripts is also stored in this folder as retFig0[4,5,8].mat and dwix[GSVS-stdhat,GSVS,MPPCA].nii. Script used to run the NLSAM method from https://github.com/samuelstjean/nlsam/releases/download/v0.6.1/nlsam_0.6.1_linux_x64.zip is also included: nlsam.run.sh.

NOTE 2: Computation times of the provided execution mode ('Quick' in flag "typExec") on an 8(16) x Intel(R) Core(TM) i7-5960X CPU @ 3.00GHz 64GB RAM with a GeForce GTX TITAN X range from 1' to 5' depending on the denoising method. Note Experiment 1 (Fig. 4) runs seven methods, Experiment 2 (Fig. 9) runs four methods, Experiment 3 (Fig. 6) runs two methods and Experiment 4 (Fig. 5) runs three methods.