GithubHelp home page GithubHelp logo

torbenelund / 2d-lookup-tools-for-mp2rage Goto Github PK

View Code? Open in Web Editor NEW
1.0 4.0 0.0 46.8 MB

Convert MP2RAGE with non-bijective UNI-T1 transfer curves to R1-maps

License: GNU General Public License v3.0

MATLAB 100.00%
mp2rage 2d-lut r1-mapping b1-tollerance b1

2d-lookup-tools-for-mp2rage's Introduction

2D-lookup-tools-for-MP2RAGE

Matlab tools for dealing with non-bijective MP2RAGE transfer curves

These matlab functions can be used to translate values in a unified MP2RAGE image into $T_1$ and $R_1$ maps despite non-bijective behavior of the MP2RAGE to $T_1$ transfer curve. The method is an extention to the 1D lookup table (1D-LUT) described in [Ref 1]. The method which is described in [Ref 2] involves the creation of a second contrast based on the two RAGE image acquisitions in the MP2RAGE acquisition. The second contrast is referred to DSR (Difference Sum Ratio) and is simply derived as $$I_{DSR}=\frac{|I_1|-|I_2|}{2(|I_1|+|I_2|)}$$

Using Bloch simulations the expected value of $I_{DSR}$ can be derived based on $T_1$ and the sequence parameters, and my minimising the (weighted) distance to the $I_{UNI},I_{DSR}$ curve a $T_1$ value can be found for any pair of measured $I_{UNI},I_{DSR}$

As seen in the figure below it is now possible to find the correct $R_1$ value, even in situations where one UNI value corresponds to two $R_1$ values. This is done by taking into account the second contrast, DSR, which also, in some situations have two $R_1$ values corresponding to one DSR value. By calculating the shortest (possibly weighted) distance bewteen a measured pair of (UNI,DSR) values and theoretical values in 2D-LUT a $R_1$ value can be derived for each (UNI,DSR) pair and an $R_1$-map can be created.

image

Installation:

Add the entire 2D-lookup-tools-for-MP2RAGE directory to your matlabpath with subdirectories, and make sure you have the required software installed.

Usage:

Start with the examples: VisualiseB1EffectsExample.m, Simple2DLUTConversionExample.m and DisplayConvertedImagesAndExplain2DLUTExample.m in that order. Or simply run the AllExamplesWithExplanation.m example to run them all. The script ScriptUsedToCreateFiguresInThePaper.m cannot be executed on the example dataset, but illustrates how the figures in [Ref 2] were created. The functions required for the method introduced in [Ref 2] are all located in the func directory. In the SPM_and_CAT12_wrappers directory a number of wrappers to the SPM12 and CAT12 software has been included. These wrappers are used e.g. to project out local curvature from the $R_1$-surfaces. The animated GIF illustrates the effect of performing this correction.

CurcatureCorrectionMovie

References:

  1. Marques JP, Kober T, Krueger G, Van Der Zwaag W, Van De Moortele PF, Gruetter R. MP2RAGE, a self bias-field corrected sequence for improved segmentation and $T_1$-mapping at high field. NeuroImage. 2010 Jan;49(2):1271–81.

  2. Ruijters L, Lund TE, Vinding MS. Tolerating B1+-inhomogeneity in MP2RAGE-based $R_1$-mapping by calculating second contrast that permits previously problematic sequence parameters. Submitted to Magnetic Resonance in Medicine. Preprint avialable on bioRXiv: https://www.biorxiv.org/content/10.1101/2024.04.10.588855v1

If you use it, please cite the paper:

Ruijters L, Lund TE, Vinding MS. Tolerating B1+-inhomogeneity in MP2RAGE-based $R_1$-mapping by calculating second contrast that permits previously problematic sequence parameters. Submitted to Magnetic Resonance in Medicine. Preprint avialable on bioRXiv: https://www.biorxiv.org/content/10.1101/2024.04.10.588855v1

Required software:

José P. Marques' MP2RAGE related scripts: https://github.com/JosePMarques/MP2RAGE-related-scripts

SPM12: https://github.com/spm/spm12

2d-lookup-tools-for-mp2rage's People

Contributors

torbenelund avatar

Stargazers

 avatar

Watchers

 avatar  avatar  avatar  avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.