3D-MSI metal artifact reduction sequences are increasingly being utilized to assess bone and soft tissue in the near vicinity of metallic implants. There are a variety of sources of remaining signal loss in 3D-MSI, some of which remain elusive to recover [1]. However, one commonly encountered and avoidable source of signal loss in 3D-MSI occurs at the edges of thin slabs prescribed tightly around large orthopedic implants. Figure 1 demonstrates this effect, using a set of 3D-MSI images of a total hip replacement at 1.5T. Coronal in-plane acquired images are shown, along with reformatted axial slab images. Both images were acquired with identical acquisition parameters (including the number of MSI spectral bins) with the exception of the number of slab encodes. In the displayed coronal slice on the posterior edge of the implant, there is clear signal loss (white arrows) in the thinner-slab acquisition (right, 6cm slab thickness). This signal is clearly recovered in the thicker lab acquisition (left, 14.4cm slab thickness), which accurately depicts the implant boundary in this slice.

The source of the signal loss visualized in Figure 1 is graphically illustrated in Figure2. Ultimately, the signal loss is an issue of spectral coverage that is compromised by the presence of the slab-selective gradient. The use of slab-selective gradients in MAVRIC SL [2] or SEMAC 3D-MSI [3] sequences superimpose an additional linear off-resonance distribution onto the native implant-induced distribution that must be covered with the 3D-MSI spectral acquisition window. Figure 2 provides frequency offset histograms across a set of slabs of varying sizes. The non-selective histogram shows a tight distribution, which is successively broadened by selective slabs of decreasing thickness. The purple line in the zoomed histogram edge indicates a frequency cutoff estimated from a 3D-MSI external calibration process [4]. When slabs of increasingly reduced thickness are prescribed, the off-resonance distribution moves well beyond the native frequency cutoff. The lack of excitation of these off-resonance components manifests as signal loss, typically at the edges of slabs.

This source of signal loss in 3D-MSI is deterministic and can be predicted, and subsequently avoided using external spectral calibration procedures [4]. Briefly, external 3D-MSI calibration requires the acquisition of a low-resolution non-selective 3D-MSI calibration scan. This calibration scan is used to generate 3D-MSI field maps [2], mask relevant field information, and determine frequency offset cutoffs using cumulative distribution function analysis of the frequency offset maps.

To predict and avoid slab-selection induced signal loss, the slab-selective gradient must be incorporated into the spectral calibration algorithm. More directly, the minimum slab thickness, at a given slab-center location, that will avoid signal loss can be computed using the cumulative distribution function analysis. Here, the minimum slab thickness is determined through an iterative process whereby the slab is reduced until the cutoff analysis indicates that signal will be missed in the 3D-MSI acquisition. Data was collected from a clinical research subject who provided written informed consent into a protocol approved by the MCW IRB.

Figure 3 displays a 3D-MSI calibration dataset on the hip arthroplasty shown in Figure 1. The calibration acquisition required ~1:20 of total acquisition time. Coronal in-plane images are shown for the magnitude data and 3D-MSI field map. Reformatted axial planes are shown to highlight the acquired and determined slab thicknesses. The aforementioned algorithm was utilized to compute a minimal slab thickness, which is displayed in the axial plane, along with the two slabs acquired for the images shown in Figure 1. The minimal slab thickness was computed to be 13.25 cm, which is just under the thick slab acquired in Figure 1 (14.4 cm). The thin slab from Figure 1 (6.0 cm) is well under this minimal threshold, which explains the signal missing from the displayed edge slice in Figure 1. The presented methods can be utilized to prospectively guide prescriptions for slab-selective 3D-MSI. With the acquisition of a relatively short calibration scan, the requisite number of spectral bins can efficiently be determined simultaneously with a minimum slab thickness that will avoid any edge signal loss. For MAVRIC SL 3D-MSI, the minimal slab-thickness to avoid edge signal loss is independent of the computed number of spectral bins, which is because the MAVRIC SL slab-gradient depends on the number of spectral bins, which introduces a “chasing condition” into the bin/slab computation. Alternatively, SEMAC can independently determine requisite spectral coverage based on the calibration and a desired spatial excitation region. It is anticipated that the presented methods can aid in the robust acquisition of high-quality 3D-MSI in routine clinical settings.