Meeting Banner
Abstract #0067

Nonlinear Formulation of the Magnetic Field to Source Relationship for Robust Quantitative Susceptibility Mapping

Tian Liu1, Cynthia Wisnieff2, 3, Min Lou4, Weiwei Chen5, Pascal Spincemaille3, Yi Wang2, 3

1MedImageMetric LLC, New York, NY, United States; 2Biomedical Engineering, Cornell University, Ithaca, NY, United States; 3Radiology, Weill Cornell Medical College, New York, NY, United States; 4Neurology, The Second Affiliated Hospital, Zhejiang University School of Medicine, Hangzhou, Zhejiang, China; 5Radiology, Tongji Hospital, Tongji Medical College, Huazhong University of Science& Technology (HUST), Wuhan, Hubei, China

Quantitative Susceptibility Mapping is becoming an increasingly active area of scientific and clinical investigations. In practical applications, there are sources of errors for QSM including noise, phase unwrapping failures and signal model inaccuracy. To improve the robustness of QSM quality, we propose a nonlinear data fitting for field map estimation and dipole inversion to reduce noise and phase unwrapping failures, and a method for model error reduction through iterative tuning. Compared to the previous linear QSM method, this nonlinear QSM method reduced checkerboard pattern in high susceptibility regions in healthy subjects and markedly reduced artifacts in patients with intracerebral hemorrhages.

Abstract PDF Presentation Video

Keywords

absence account accounted acquisition adjusted affiliated anatomical artifact artifacts assessed assuming bandwidth bias biomedical blue catch china circles college combined complex components conjugate considering constraint cross decay decomposition decreased degree denotes develop developed deviation diagonal dipole engineering error errors estimation exemplary expected exponential failed fidelity field fitted fitting fittings formulation frequencies frequency freshly functions generated gradient hemorrhage hemorrhages hospital identified imaginary implementation improvement initial instead inversion inverted iteration iterative keeping known linear linearized magnitude manner mapping marked markedly matrix measured medical medicine merit minimizing model neither neurology noise none nonlinear numerical operator part pascal patient patients people press previously prior problem problems proper properly propose proposed published quality quantitative radiology rank reconstructed reconstruction reduced reduction reflecting relationship reliability reliable repeated repetitions republic residual respectively robust school score scored scores shadowing signed simulated simulation slice solve solved solving source square structural subject susceptibility technical term termed theory treatment true tuning unsolved unwrapping