hide
Free keywords:
-
Abstract:
Purpose of the Software: Realistic simulations of CEST effects are very important with regard to e.g. the choice of saturation parameters or fitting of experimental results. However, most of the time only very few CEST pools or even single pools are used for simulations with regard to computation time and code simplicity. In this work, we present a fast, MATLAB compatible, toolbox for Bloch-McConnell simulations, which is designed in a flexible way that allows simulations of a system with arbitrary number of CEST pools and all types of pre-saturation. Methods/Implementation: The source code is written in C?? and is callable from MATLAB as a mex function. openMP is used for parallelization of the different Z-spectra offset simulations and eigen1 for linear algebra operations. The dynamic design allows a variable number of CEST Pools and an optional MT pool with either a Lorentzian or a Super-Lorentzian line-shape. A simple 3 pool model e.g. is defined by the relaxation rates R1 and R2, the exchange rate k, proton fraction f and the chemical shift dx of a CEST and an MT Pool. In addition, the MT line-shape and the relaxation rates of the water pool need to be set. Figure 1 shows an exemplary Z-Spectrum with 15 CEST pools in total (9 in vivo pools2 and 6 paraCEST pools3) and a Super-Lorentzian shaped MT Pool. The code was compiled for a 64-bit Windows 10 OS, using Visual Studio 2017 Professional. Simulation parameters for continuous wave saturation were: B0: 18.8 T; T1: 4 s; T2: 1 s; 1 rectangular pulse with tpulse = 1 s; B1: 15 lT; Dx = ± 700 ppm; 30 k Z-spectrum samples. Using 6 openMP threads, the computation time was approx. 13 s on an Intel i7-7700 K Kaby Lake CPU. The same setting, but with just one CEST and a Lorentzian shaped MT Pool was solved in approx. 0.26 s. Figure 2 shows the simulation of the parameters from reference 2 at B0 = 9.4T; T1 = 2 s and T2 = 35 ms. Here, pulsed CEST with a train of 150 15 ms Gaussian pulses with 50 pulse samples and a duty cycle of 50% was simulated for 100 Z-spectra samples between ± 4.5 ppm. For the 9 CEST Pools and an MT pool this resulted in a computation time of approx. 20 s. The same setting with only water and the MT pool is solved in * 1.1 s and in * 0.25 s for only the water baseline.
