A robust MRI water‐fat separation algorithm based on field map and fat fraction map smoothness

作者：Kecheng Yuan, Qingyun Liu, Xuhe Huangfu, Jiaen Wang, Rong Zhang, Penghui Luo, Changliang Wang, Fulang Qi, Chuang Zhang, Lin Chen, Bensheng Qiu · 发表于：Medical Physics · 年份：2025 · DOI：10.1002/mp.70101 · 被引用次数：1 · 研究领域：Advanced MRI Techniques and Applications、NMR spectroscopy and applications、Advanced Neuroimaging Techniques and Applications

BACKGROUND: Magnetic resonance water-fat separation is an important imaging technique for distinguishing water and fat signals, enabling accurate tissue characterization and fat quantification in both clinical and research settings. However, achieving robust separation remains challenging, especially in the presence of complex background noise and rapidly varying magnetic field inhomogeneities. PURPOSE: To improve robustness in challenging acquisition scenarios, including low field strength systems and large field of view (FOV) imaging with complex boundary regions, by leveraging the inherent smoothness of field maps and fat fraction maps. METHODS: * decay correction, and field inhomogeneity compensation. The signal model accounted for chemical shift and relaxation effects, requiring at least three uniformly spaced echoes. Nonlinear parameters were estimated via a two-stage optimization framework, with water and fat amplitudes derived using a least-squares solution. To further improve robustness, we leveraged the spatial smoothness of both the field map and fat fraction map. Erroneous voxels were identified by local field discontinuities and refined through Local Polynomial Surface Fitting (LPSF) applied to neighboring field and fat fraction values. A final field value was selected based on agreement between smoothed and candidate fat fraction estimates within a narrow search window. RESULTS: Validation on the ISMRM 2012 Challenge dataset, an agar-based water-fat phantom data...