Open MPI data in python code

[obaida90]

I'm new in the field of MPI. I want to open the Open MPI data using the python code for the system matrix (SM) reconstruction given in the website. My issue is how I can choose the SM and measurement pairs to run the code.

[hofmannmartin]

Suitable pairs must be measured with the matching excitation sequence and tracer. Information about both can be found here for all datasets.

As a side note, I would advise you to start with datasets measured with a 2D Lissajous excitation, since the corresponding system matrices are much smaller than for a 3D Lissajous excitation.

[obaida90]

Thank you for reply
I would like to open the data for example the first 1D system matrix with the 1D concentration phantom measurement in Jupyter notebook, although they have the same tracer and sequence but there was an error sayed "cannot reshape array of size 112632000 into shape (2,1083000)" ?

[hofmannmartin]

There are numerous challenges associated with using MPI data, and your question is a bit vague. Can you please provide a code sample to illustrate your problem?

[obaida90]

Yes offcourse, the code is here: https://github.com/MagneticParticleImaging/MDF/tree/master/python ; what I did is I combined the three codes in this link together (KaczmarzReg.py, pseudoinverse.py and reco.py).

[hofmannmartin]

Could you narrow down, in which file and at line the Error occurs?

[obaida90]

Ok in fact I faced several errors but the most common one is in these lines :
1)# merge frequency and receive channel dimensions
S = reshape(S, (shape(S)[0]*shape(S)[1],shape(S)[2]))
u = reshape(u, (shape(u)[0],shape(u)[1]*shape(u)[2]))

2. reconstruct using signular value decomposition

U, Sigm, V = svd(S, full_matrices=False)
csvd = pseudoinverse(U, Sigm, V, u, 5e2, True, True)

[chuyuenin]

I found the OpenMPI data set link at https://media.tuhh.de/ibi/openMPIData/data/, but the system matrix and the measured signal must match each other to reconstruct the image, but these two folders (one system matrix folder, the other is the folder for the measured signals) I don't know which system matrix matches which measurement data. I mainly want 2D data

[tknopp]

just look here: https://magneticparticleimaging.github.io/OpenMPIData.jl/latest/datasets.html
you need to match tracer and sequence. then they will be compatible.

[chuyuenin]

Thank you for your reply. I browsed the URL you provided (https://magneticparticleimaging.github.io/OpenMPIData.jl/latest/datasets.html) and found the 2D system matrix data as shown in Figure 1, but I found it in your The OpenMPIData dataset (https://media.tuhh.de/ibi/openMPIData/data/calibrations/) found a lot of system matrix files, as shown in Figure 2, can you tell me which one is the 2D system matrix?

[mgraeser]

You can use all 2D Lissajous system matrices, so calibrations number 2.mdf and 5.mdf in the table sent by Tobias. Any experiment with number 2.mdf (2D Lissajous) will be compatible.

[sesafe]

I am very interested in your paper titled "3d-SMRnet: Achieving a new quality of MPI system matrix recovery by deep learning", but when I want to run the code you provided, I found that the data set link you provided is invalid. . Can you provide a link to the dataset?

[mgraeser]

Under Datasets https://magneticparticleimaging.github.io/OpenMPIData.jl/latest/datasets.htmlYou should find every information regarding grid and FOV size. Or do you mean the dimensions of the matrices in the MDF. If so please have a look at the MPI Data Format Paper MDF: Magnetic Particle Imaging Data Format https://arxiv.org/abs/1602.06072

[chuyuenin]

Hello, can you help me explain (https://media.tuhh.de/ibi/openMPIData/data/calibrations/) the files 7.mdf to 17.mdf in the URL are the system matrices of several dimensions. And which are the high-resolution system matrices and which are the corresponding third-resolution system matrices, as shown in the figure.

[sesafe]

Under Datasets https://magneticparticleimaging.github.io/OpenMPIData.jl/latest/datasets.htmlYou should find every information regarding grid and FOV size. Or do you mean the dimensions of the matrices in the MDF. If so please have a look at the MPI Data Format Paper MDF: Magnetic Particle Imaging Data Format https://arxiv.org/abs/1602.06072

Hello! When I reproduced the code provided in the paper "3d-SMRnet: Achieving a new quality of MPI system matrix recovery by deep learning" on github, I found that there was a problem with the data set link and could not be jumped. Then the dataset address you provided is a .mdf file, but the code requires a .h5 file. Could you please provide the data set of the original code? Thank you.

[chuyuenin]

Hello, may I ask (https://media.tuhh.de/ibi/openMPIData/data/) that the OpenMP data under this link has two folders: system matrix and voltage signal. Please tell me which 3D system matrix and voltage signal match each other to achieve imaging.