Fat-Water Separation Using a Region-Growing Algorithm With Self-Feeding Phasor Estimation

Purpose: To develop a novel region-growing algorithm with self-feeding phasor estimation for robust fat-water separation. Theory and Methods: The proposed seed pixel identification and region-growing methods were performed independently at different resolutions. Multiple phasor maps were obtained at lower resolutions and then merged into a new seed map, which was used to generate the final phasor map at the finest resolution. The final fat and water images were reconstructed based on this phasor map. The proposed method was compared with traditional region-growing methods, multiresolution methods, and graph-cut methods using data from the ISMRM 2012 Challenge. All methods were scored on a scale of 0 to 10000. Results: The average score of all 17 data sets from the ISMRM 2012 Challenge was 9928, with 13 of the 17 scores surpassing 9900. The lowest score was 9697 from data set # 12; there was no apparent fat-water swap observed throughout these data sets. Conclusions: The self-feeding mechanism of phasor estimation ensures the reliability of seed pixel selection at the finest resolution. Compared with traditional multiple resolution methods and region-growing methods, the proposed method is shown to be more robust when applied to disjoint areas and to regions with strong field inhomogeneity. (C) 2016 International Society for Magnetic Resonance in Medicine