MRI $RF$ pulse simulation in MALTAB

This repository is a MATLAB application that simulate the response of MRI RadioFrequency (RF) pulses. The app is a GUI, and the code also made to be used purely programmatically.

1. Open the GUI app
2. Click on a pulse in the library list.
3. The selected $RF$ pulse is loaded with default parameters, plotted in the GUI, and it's simulation triggered. The simulation is plotted automatically : magnetization vector across time $M_{xzy}(t)$, slice profile $\Delta Z$, chemical shift profile $\Delta B_0$.

The application is highly object oriented, to take advandtage of heritage and composition of several abstract classes.

Also, you can use your own pulses in the app by :

• A super fast method : fill the $RF$ pulse shape ($B1$ curve $GZ$ curve) in the the USER_DEFINED pulse. This is "empty" pulse, used as placeholder in the app. Example usage : if you have Brucker pulse file (.exc .rfc .inv), load it with load_bruker_RFpulse, fill the USER_DEFINED pulse, then trigger the GUI so simulate the profile.
• An ergonomic method made for interativity : add your own $RF$ pulse class in +mri_rf_pulse_sim/+rf_pulse/+local/ so it will appear in the GUI library. This directory is not versioned in this repo

Features

GUI

The GUI have 3 independent panels :

• Pulse definition : It shows the library of pulses, and the selected pulse, including its shape and the UI parameters.
• Simulation parameters : You define the range and granularity (number of points) for the slice profile evaluation $\Delta Z$ and the chemical shift $\Delta B_0$ evaluation.
• Simulation results : Displays $M_{xzy}(t)$, the slice profile $\Delta Z$, and the chemical shift $\Delta B_0$ profile.

Scripting

Here is some examples of non-GUI analysis :

• Why SINC is used for slice selection instead of to RECT ?
• Too much B1max ? RF clip ? maybe increase pulse duration
• FOCI is derivezd from HS pulse. But is it better ?
• Why do we need a slice selection gradient rewinder ?

Object oriented programming

All pulses are objects.
Pulses can inherit from others : FOCI is derived from HyperbolicSecant.
Pulses can be composed of several abstract classes. For example, slr_mb_verse is a SLR base waveform, then the MultiBand algorithm is applied to excite several slices, and finnally the VERSE algorithm reduces it's duration and $B1_{max}$ using constrains.

Re-usability

One of the objectives here is to centralize the equations/algorithms of $RF$ pulse so they can be almost copy-pasted in other programming environments, like a complete sequence simulator, or a sequence development environment from your manufacturer.
One difficulty when looking in the literature is that different sources can have different vocabulary or different parameters. A typical example is the HyperbolicSecant, which is the extremely well described, but with a large variety of implementation using different input parameters.

Examples

HS : Hyperbolic Secant

FOCI : Frequency Offset Corrected Inversion

Download and install

1. Clone the repository with
   • git clone --recurse-submodules https://github.com/benoitberanger/mri_rf_pulse_simulation_matlab.git
2. In Matlab, cd /path/to/mri_rf_pulse_simulation_matlab
3. Start the app with mri_rf_pulse_sim.app()

Limitations

• MATLAB R2023a+ ? maybe few release earlier, but I did not test them.

External dependency

None, except for :

• For SLR pulses :
  • DSP System Toolbox
  • Signal Processing Toolbox

Tested on

MATLAB R2023a+

Alternatives

In all alternatives that I found, in Python, MATLAB, Julia, all of them have very nice features, but none has the same interactivty and ergonomy.

• Python : https://github.com/mikgroup/sigpy
• Matlab : https://github.com/leoliuf/MRiLab
• Julia : https://github.com/cncastillo/KomaMRI.jl

TODO list

https://github.com/users/benoitberanger/projects/2