Linear projection-based CEST reconstruction

Demo codes for linear projection-based CEST reconstruction and L1-regularization-based feature selection ('CEST-LASSO'). Uses FISTA as submodule for solving LASSO problems, so use git clone --recursive

• linearCEST_demo1_pinv.m: Demo of fitting linear regression coefficients to a training dataset that directly map from raw, uncorrected Z-spectra to target contrast parameters of interest, which are here conventionally fitted Lorentzian parameters describing APT, NOE, MT and amine effects. Only uses Matlab's native pinv function.
• linearCEST_demo2_LASSO.m: Example for L1-regularization-based feature selection ('CEST-LASSO') for the linear CEST reconstruction method. The regularization enforces sparsity of Z-spectral offsets, such that a potential acceleration of CEST measurements by acquiring less offsets can be achieved. Uses the external FISTA repo for solving LASSO objectives.

Demo data can be downloaded from here.

---

Felix Glang^1*, Moritz S. Fabian², Alexander German², Katrin M. Khakzar², Angelika Mennecke², Andrzej Liebert³, Kai Herz^1,4, Patrick Liebig⁵, Burkhard S. Kasper⁶, Manuel Schmidt², Armin M. Nagel³, Frederik B. Laun³, Arnd Dörfler², Klaus Scheffler^1,3, Moritz Zaiss^1,2

¹Magnetic Resonance Center, Max Planck Institute for Biological Cybernetics, Tübingen, Germany
²Department of Neuroradiology, University Hospital Erlangen, Erlangen, Germany
³Institute of Radiology, University Hospital Erlangen, Friedrich-Alexander-Universität Erlangen-Nürnberg (FAU), Germany
⁴Department of Biomedical Magnetic Resonance, Eberhard Karls University Tübingen, Tübingen, Germany
⁵Siemens Healthcare GmbH, Erlangen, Germany
⁶Department of Neurology, University Clinic of Friedrich Alexander University Erlangen-Nürnberg, Erlangen, Germany

Correspondence:
Felix Glang
Magnetic Resonance Center
Max Planck Institute for Biological Cybernetics
Tübingen, Germany
felix.glang@tuebingen.mpg.de