Accelerating sequences in the presence of metal by exploiting the spatial distribution of off-resonance

Abstract

Purpose: To demonstrate feasibility of exploiting the spatial distribution of off-resonance surrounding metallic implants for accelerating multispectral imaging techniques.

Theory and methods: Multispectral imaging (MSI) techniques perform time-consuming independent three-dimensional acquisitions with varying radio frequency offsets to address the extreme off-resonance from metallic implants. Each off-resonance bin provides a unique spatial sensitivity that is analogous to the sensitivity of a receiver coil and, therefore, provides a unique opportunity for acceleration. Fully sampled MSI was performed to demonstrate retrospective acceleration. A uniform sampling pattern across off-resonance bins was compared with several adaptive sampling strategies using a total hip replacement phantom. Monte Carlo simulations were performed to compare noise propagation of two of these strategies. With a total knee replacement phantom, positive and negative off-resonance bins were strategically sampled with respect to the B0 field to minimize aliasing. Reconstructions were performed with a parallel imaging framework to demonstrate retrospective acceleration.

Results: An adaptive sampling scheme dramatically improved reconstruction quality, which was supported by the noise propagation analysis. Independent acceleration of negative and positive off-resonance bins demonstrated reduced overlapping of aliased signal to improve the reconstruction.

Conclusion: This work presents the feasibility of acceleration in the presence of metal by exploiting the spatial sensitivities of off-resonance bins.

Keywords: acceleration; imaging near metal; multispectral imaging; off-resonance.

Figure 1

The large susceptibility difference of metal induces spatially dependent off-resonance in adjacent tissue. (a) A digital model was generated with a 3D laser scanner of a hip implant (CoCr/Mo head, Ti femoral stem). The off-resonance fieldmap is calculated with a Fourier based k-space filtering (19) and demonstrates the characteristic dipole pattern and extreme off-resonance. Frequency offsets adjacent to the implant exceed 20 kHz, although the colormap was chosen to visualize the dipole effect. (b) Multispectral imaging (MSI) techniques acquire multiple 3D acquisitions of varying frequency offsets. These off-resonance bins represent spatially unique portions of the composite image.

Figure 2

A metallic object provides unique opportunities for acceleration of MR imaging. (a) The calculated fieldmap from a titanium sphere illustrates the dipole effect and spatial constraints of the positive and negative components of off-resonance. These signed components are dependent on the geometry of the object in relation to the direction of the B₀ field. (b) Sampling strategies can exploit this knowledge to minimize overlapping aliased signal. Simulated signal from off-resonance bins centered at -3375 Hz and +3375 Hz and a full-width at half max of 1125 Hz are shown with several undersampling strategies. Undersampling with a reduction factor of 3 in a direction perpendicular to B₀ ( R_⊥ = 3) causes overlap of aliased signal only in the negative off-resonance bin (white arrows). If the undersampling is instead applied in a direction parallel to B₀ ( R_∥ = 3), overlap of the aliased signal only appears in the positive off-resonance bin. The sampling flexibility of ORE enables independent acceleration of the negative ( R^-) and positive ( R⁺) off-resonance components of the signal in order to minimize signal aliasing.

Figure 3

Optimal kernel calibration will account for unique sampling patterns in the off-resonance dimension. (a) Conventional k-space-based parallel imaging reconstruction algorithms optimize kernel calibration with each unique sampling pattern. (b) ORE reconstruction accounts for the independent sampling pattern of each off-resonance bin.

Figure 4

The independent acquisitions of MSI methods enable an adaptive off-resonance reduction factor that can be exploited to preserve SNR, exploit the sparsity of the far-off-resonance bins, and optimize reconstruction accuracy. Scheme 1 uses a uniform undersampling pattern ( R = 1.5) across all off-resonance bins. Adaptive schemes 2-4 fully sample the three central bins and have increasing accelerations for far off-resonance bins. The y-axis corresponds to the outer k-space acceleration. The effective reduction factor, R_eff , includes acquisition of the ACS region (40 k_x lines).

Figure 5

An adaptive sampling pattern across off-resonance bins dramatically improves reconstruction compared to a zero-filled reconstruction. ORE reconstruction of a uniform sampling pattern across off-resonance bins contains residual aliasing emphasized in the error map. An adaptive sampling pattern fully samples the central off-resonance bins and accelerates the sparse far-off-resonance bins. Aliasing of far-off-resonance is visible in the zero-filled images using adaptive sampling. This approach results in highly accurate reconstructions compared to uniform sampling. The ACS data used for calibration was not included in final reconstruction.

Figure 6

ORE reconstruction effectively removes signal aliasing. The adaptive sampling scheme 3 from Figure 4 accelerates all but the central three bins, causing aliasing in the direction perpendicular to B₀ seen in the zero-filled images. The aliased signal from each off-resonance bin is removed using ORE reconstruction. Only small residual aliasing from ORE is present in the error map from the far off-resonance bins that contain signal directly adjacent to the implant. The ACS data used for calibration was not included in this reconstruction.

Figure 7

Noise maps highlight the noise enhancement in undersampled areas without adequate spatial sensitivity variation using a uniform sampling pattern. An adaptive off-resonance acceleration scheme has relatively little noise enhancement and is contained to regions of far off-resonance that were undersampled.

Figure 8

Multidimensional ORE acceleration can control the aliasing to improve reconstruction. Sampling can exploit the known dipole nature of the off-resonance. Undersampling the negative off-resonance bins in a direction parallel to B₀ ( R^-_∥ = 3) and the positive off-resonance bins in a direction perpendicular to B₀ ( R⁺_⊥ = 3) provides the least amount of aliased signal overlap and displays an ORE reconstruction with the least error. The ACS was not included in the zero-filled images but was included in the composite images.