Guidewires are essential tools for intravascular interventions – unfortunately, conventional metallic guidewires can cause excessive heating during MRI procedures.¹ Recently, MR-safe guidewires have been developed from a glass fiber reinforced structure with embedded iron particles as local passive markers. The susceptibility artifacts of these markers depend strongly on the echo time (TE) and on the orientation against B₀.² Recently, a radial acquisition scheme with variable TEs was proposed yielding a more homogeneous artifact.³ In this work a simulation framework was implemented to gain insight into the artifact behavior and, thus, to further optimize radial acquisition schemes.

Simulation Framework

In the simulation framework the intra-voxel dephasing of a magnetic distortion field B_z’ (e.g. from the markers of the guidewire) is calculated. MR signals are computed from the magnetization on a 15-fold denser ‘spin’-grid in each spatial direction compared to the intended imaging resolution. For each grid location, the magnetization evolution during data acquisition is simulated, and all grid values are finally summed up to form the simulated k-space. Signal values of ‘spins’ which are spatially located within the guidewire material or which are spectrally located outside the range of the slice-selection process due to influence from B_z’ are set to 0 while the remaining ones are initialized as 1. The simulation is performed for conventional Cartesian sampling as well as for radial acquisitions with constant and variable TEs.

To simulate the artifacts of an MR-safe guidewire (MaRVis Medical GmbH, Hannover, Germany), the guidewire was modeled as a chain of spherical iron particles (size: 8±2µm) placed along its central strand (mean particle distance along guidewire direction: 50±30µm, radius of central strand: 60µm). Particle sizes and distances along the guidewire followed a Gaussian random distribution whereas the axial variations of the particle position (perpendicular to guidewire orientation) followed a uniform random distribution. Each iron particle was represented by a point-like dipole field so that the total B_z’ can be calculated from the summation of individual dipole fields:

$$B_z^{~'}=\sum_i\frac{\mu_0m_z}{4\pi}\cdot\frac{2(z-z_i)^2-(x-x_i)^2-(y-y_i)^2}{((x-x_i)^2+(y-y_i)^2+(z-z_i)^2)^{5/2}}$$

Here, (x_i,y_i,z_i) reflects the position of the i^th dipole. The magnetic moment m_z of the particles is assumed to be aligned with B₀ and cal­cu­lated from the particle volume and the sa­turation magnetization of iron.

Variable-TE Acquisition Scheme

In the radial acquisition with variable TEs, a different TE_n is used for every spoke n so that TE is shortest (TE_min=1.8ms) for readout-direction parallel to B₀ and longest (TE_max=10.8ms) perpendicular to B₀ to minimize the directional anisotropy of the dipole artifact. For all other directions, two different variation schemes are implemented: (1) Sine-modulation ³: $$$\mathrm{TE}_n=\mathrm{TE}_\mathrm{min}+(\mathrm{TE}_\mathrm{max}-\mathrm{TE}_\mathrm{min})\cdot\vert\mathrm{sin}(\phi_n)\vert$$$, and (2) Sine-squared modulation: $$$\mathrm{TE}_n=\mathrm{TE}_\mathrm{min}+(\mathrm{TE}_\mathrm{max}-\mathrm{TE}_\mathrm{min})\cdot\mathrm{sin}^2(\phi_n)$$$. The latter scheme is chosen as the inverse of the cos²-modulated dipole-field.

For both schemes, simulations were carried out with the following parameters: FOV=270x270mm², object-size=180x250mm², BW=610Hz/px, slice-thickness=5mm, matrix=128x128. The acquisition was simulated for 201 radial spokes (angular range: 2π) with 256 readout points per spoke (2-fold oversampling). Images were reconstructed using the NUFFT algorithm.⁴ Simulations were based on a GPU implementation ⁵ and remaining computation was done in MATLAB.

For comparison, images of the guidewire were acquired at 1.5T (Siemens Symphony) using the same parameters as in the simulation (TR=14ms, α=15°). Both, simulations and measurements, were also carried out using a conventional Cartesian acquisition as reference. To measure the artifact width, profiles perpendicular to the wire were defined and the FWHM was computed for sections of the wire perpendicular and parallel to B₀ (Fig. 2).

Measured and simulated Cartesian (Fig. 1a,e) and radial (Figs. 1b,f) images consistently show the expected anisotropic artifact for constant TE and a more uniform artifact pattern for the variable-TE imaging strategies (Figs. 1c,d,g,h). Quantitatively, simulated artifact widths were systematically smaller (15-25%) whereas the artifact ratios (r_w) are in agreement which means that our simulations correctly describe the artifact behavior. Only minor differences in r_w were found for the sin- and sin²-modulated acquisitions, although the artifact visually appears more defined in the sin²-case.

Even though the simulation systematically underestimates the absolute artifact width, simulations proved as a useful tool for sequence optimization as reflected in a correct reproduction of the artifact ratios. To further improve the simulation, iron particles could be simulated more realistically, e.g. as cluster-forming ellipsoids. Based on the simulations, further optimizations such as different TE modulation schemes, variation of TE_max and TE_min, and variable-TRs can be easily tested. This will facilitate the development of efficient data acquisition strategies for improved guidewire visualization.