Quantitative susceptibility mapping of the knee: A comparison of approaches for addressing fatty tissue

Sammanfattning: Background/purpose Quantitative susceptibility mapping (QSM) is an MRI-based post processing technique to evaluate the magnetic susceptibility, χ, of human tissue using phase data. While brain imaging has been the most common application, other areas have also raised interest. For example, QSM of the articular cartilage of the knee has been suggested as a means to study the degeneration of this tissue. This would be of interest, as degeneration of articular cartilage is strongly related to the progression of osteoarthritis (OA). However, QSM of the knee is not straightforward, as the signal from fatty tissue complicates the reconstruction process due to its many frequency components. To avoid this issue, masking has been used to exclude fatty tissue from the reconstruction process. This is a relatively simple approach to a complicated problem, but, it might also be a source of bias in the susceptibility maps obtained. Another approach is the removal of chemical shift through chemical shift encoded imaging (CSEI) before QSM. The purpose of this work was to evaluate and compare the effects these approaches have on the estimated susceptibility values. Materials and methods Numerical phantoms were created based on two sets of MR images of knees, by defining various lean and fatty tissue compartments and assigning these literature susceptibility values. From these, phase images were simulated, which, in turn were used for QSM reconstruction. Comparisons were conducted between different background field removal techniques, and masking alternatives excluding: 1) no tissue, 2) bone marrow and 3) all fatty tissues from the reconstruction process. The performance of the different alternatives was evaluated by how well the ground truth was reproduced. Potential sources of error in the CSEI were also simulated. The two approaches to handle signal from fatty tissue in the susceptibility calculation were also tested in vivo. Here, comparisons were made using the same background field removal techniques and masking alternatives as for the simulations. Results In the simulations, the ground truth was reproduced most reliably when no tissue was excluded from the reconstruction process, while masking of fatty tissue introduced a bias in the obtained susceptibility values. The size of this bias depended on the background field removal technique employed, how much of the fatty tissue that was excluded and the geometry of the knee evaluated. Results obtained from in vivo data corresponded to results reported in earlier studies and confirmed the simulation results of this study. However, both with and without CSEI, artefacts were seen when masking of fatty tissue was not performed. These were mitigated by exclusion of fatty tissue. The severity of these artefacts was greater without CSEI. Conclusion Exclusion of fatty tissue resulted in bias in the calculated susceptibility maps. Inclusion of fatty tissue in the reconstruction processes yielded more robust results between reconstruction techniques and geometry. However, in the in vivo results, the exclusion of fatty tissue from the reconstruction process mitigated artefacts stemming from imperfect CSEI. Inclusion of fatty tissue in QSM reconstruction is advantageous, assuming that an accurate enough CSEI may be performed. If fatty tissue is excluded from QSM reconstruction, a suitable choice of background field removal technique may limit the bias caused.