Analysis of different phase unwrapping methods to optimize quantitative susceptibility mapping in the abdomen

Quantitative susceptibility mapping (QSM) is heavily impacted by phase processing of gradient echo imaging data. So far, phase unwrapping algorithms have mostly been developed and tested for neuroimaging applications. In this work, a numerical human abdomen phantom was created and used to assess the feasibility of different phase unwrapping algorithms in abdominal QSM. Furthermore, in vivo data were acquired to evaluate consistency with the simulations.

Laplacian-based, quality-guided region growing and graph-cuts unwrapping techniques were evaluated using the numerical phantom as well as an in vivo measurement. As a quality metric, root mean square error (RMSE) was calculated in order to analyze the performance of the examined unwrapping algorithms. Subsequently, susceptibility maps were generated from the resulting phase maps and compared to the ground truth. The evaluation was carried out on the whole phantom as well as individual organs.

Graph-cuts led to the most accurate and robust results among the investigated unwrapping methods. The other algorithms showed severe errors in regions with large susceptibility changes (i.e., around the lungs). Deviations from the ground-truth susceptibility were higher than in the previous brain simulations for all tested algorithms.

Graph-cuts-based unwrapping algorithms should be preferred in QSM studies in the human abdomen, where large susceptibility changes occur. For further improvement of QSM studies, unwrapping algorithms should be optimized for abdominal applications.