Background field removal technique based on non-regularized variable kernels sophisticated harmonic artifact reduction for phase data for quantitative susceptibility mapping

Publication date: October 2018
Source:Magnetic Resonance Imaging, Volume 52
Author(s): Hirohito Kan, Nobuyuki Arai, Masahiro Takizawa, Kazuyoshi Omori, Harumasa Kasai, Hiroshi Kunitomo, Yasujiro Hirose, Yuta Shibamoto
PurposeWe developed a non-regularized, variable kernel, sophisticated harmonic artifact reduction for phase data (NR-VSHARP) method to accurately estimate local tissue fields without regularization for quantitative susceptibility mapping (QSM). We then used a digital brain phantom to evaluate the accuracy of the NR-VSHARP method, and compared it with the VSHARP and iterative spherical mean value (iSMV) methods through in vivo human brain experiments.Materials and methodsOur proposed NR-VSHARP method, which uses variable spherical mean value (SMV) kernels, minimizes L2 norms only within the volume of interest to reduce phase errors and save cortical information without regularization. In a numerical phantom study, relative local field and susceptibility map errors were determined using NR-VSHARP, VSHARP, and iSMV. Additionally, various background field elimination methods were used to image the human brain.ResultsIn a numerical phantom study, the use of NR-VSHARP considerably reduced the relative local field and susceptibility map errors throughout a digital whole brain phantom, compared with VSHARP and iSMV. In the in vivo experiment, the NR-VSHARP-estimated local field could sufficiently achieve minimal boundary losses and phase error suppression throughout the brain. Moreover, the susceptibility map generated using NR-VSHARP minimized the occurrence of streaking artifacts caused by insufficient background field removal.ConclusionOur proposed NR-VSHARP method yields minimal boundary losses and highly precise phase data. Our results suggest that this technique may facilitate high-quality QSM.