In 2013, over 340,000 total hip replacements (THA) were installed in the US [1]. THA has a failure rate of 6% and 13% at 5 and 10-year benchmarks [2], which often may be related to metallic and polyethylene debris deposition from implant components. Implanted cobalt chromium metallic components and their associated debris particles have a strong paramagnetic magnetic susceptibility relative to biological materials. In cases of symptomatic THA, it is sometimes possible to identify debris from magnitude MRI data. However, in such cases it is also clinically advantageous to differentiate metallic debris from polymeric debris given the more serious tissue necrosis associated with metallic debris. Previously, we demonstrated a mechanism to utilize magnetic field maps produced with intermediate data provided by 3D-MSI artifact-reduction sequences [3] to qualitatively highlight cobalt-chromium debris deposits in vivo [4]. Here, we extend this work to provide a quantifiable regional metallosis metric, reported here as the 'mScore'. In a cohort of 15 subjects undergoing total hip revision surgery, regional mScores were computed using pre-operative 3D-MSI imaging data and then correlated with histological metallosis scores from local tissue samples retrieved during surgery.

The algorithm utilized to compute mScores from MRI data is provided in Figure 1. Briefly, a region of interest is identified in the 3D-MSI MRI. Intermediate 3D-MSI spectral data is then used to identify tissue (i.e., non-implant or noise voxels) and construct magnetic field maps across these voxels (which are spectral-bin intensity-based field map estimates). These maps are regionally processed using a background field removal technique, which exposes the local tissue-specific magnetic field [5]. Clusters of field offsets are then identified at N levels of Larmor frequency offsets (300 350 400 450 500) Hz and M cluster volume thresholds (60,120,240,360,480) mm³. The volume of identified clusters at these settings forms an NxM matrix. After applying an exponential weighting to the matrix elements (which can be used to tighten mScores for a wide spectrum of metallosis severity), the elements are summed to form the mScore.

Following IRB approval with informed written consent, 15 patients were enrolled in our study and tissue samples (~1cm³) were extracted during revision THA surgery from regions of suspected particulate debris based on pre-operative MRI. Histological scoring was performed on these samples using the Fujishiro metal particle score, which ranges from 0 (no metal particles) to 4 (significant metal particles). Due to the uncertainty of precise sample locations extracted during surgery, a relatively large volume (120 cm³) surrounding the indented extraction point was utilized for mScore analysis.

Figure 2 presents a magnitude 3D-MSI (MAVRIC SL) image around a THA. The arrow indicates the region where a tissue sample was extracted. A 3D-MSI based field map is also shown in Figure 2, which is dominated by a dipolar field signature from the cobalt-chromium femoral head component. Figure 3 depicts a zoomed image of this map in the identified region of interest, along with a background field-suppressed residual tissue off-resonance map. Figure 4 presents the residual tissue off-resonance map, along with a sample identified cluster at the 300 Hz / 250 mm³ threshold setting. The white arrows in the magnitude image and tissue field indicate a hypointense region in the magnitude image which is suspected of being a central focus of the metallic debris. A volumetric surface rendering of the implant region (blue) and the metallosis cluster is also shown. The case presented in Figures 2-4 resulted in a computed mScore of 21 and had a Fujishiro metallosis histology score of 4/4. Figure 5 displays mScore vs histology analysis for Fujishiro grades 0 and 4. A significant difference of mScore was found between the two groups (p<0.025, Wilcox-Rank-Sum test) . tight clustering of low mScores, while the Fushishiro 4/4 show a much broader spread of scores, with a much higher mean. A limitation of the study is the challenge of registering a small tissue sample from surgery to its precise location on the MRI images, which is a potential reason for the broad distribution of mScores within the Fujishiro groups. It is possible that metallosis pockets were detected in the larger mScore analysis volumes and were not biopsied at revision surgery. In addition, the mScores will inherently show more variance than the tissue samples due to the more quantitative and widespread analyses that are feasible using the full MRI dataset. This strong statistical correlation (r=0.55, p=0.035, Spearman non-parametric) demonstrates that the presented methods offer a promising potential MRI-based biomarker for metallosis assessment near total hip arthroplasty.