Abstract
PURPOSE: To implement, optimize, and validate parallel imaging (PI)-accelerated, 2D, flip angle modulated (FAM) chemical shift-encoded quantification of liver proton-density fat fraction (PDFF), with motion insensitivity. METHODS: The optimization cost function that determines flip angles in FAM was generalized for PI. Phantom studies and prospective studies in volunteers with varying liver fat levels were performed. Free-breathing FAM was acquired in the axial, sagittal, and coronal planes, with varying nominal PI acceleration factors (R) of 1.0 to 3.0. A breath-held, commercially available 3D chemical shift-encoded method was acquired as reference for PDFF. Overall image quality, qualitative SNR, and motion artifacts for all methods were Likert-scale rated. PDFF measured by FAM was compared to reference to assess bias. Test-retest repeatability was assessed for all methods by repeating acquisitions after volunteer repositioning. Noise performance was assessed with standard deviation of PDFF maps as R increased. RESULTS: The reader study (N = 3 readers/10 subjects) demonstrated excellent image quality for FAM during free-breathing, with reduced motion artifacts compared to breath-held reference (p < 0.01). PI-accelerated FAM shows fewer motion artifacts than unaccelerated FAM (p < 0.01). In all planes and accelerations, PDFF measured by FAM showed good agreement with reference PDFF measurements (mean bias: -0.4% to 2.0% PDFF; 95% limits of agreement: 2.8% to 4.0% PDFF). FAM in axial and coronal planes showed similar or improved repeatability (repeatability coefficient = 1.7% to 2.6% PDFF) compared to the reference (2.7%). Sagittal FAM shows similar or worse repeatability (repeatability coefficient = 3.0% to 3.6%). FAM with R = 2.0 has good noise performance and high SNR efficiency. CONCLUSION: FAM, in axial or coronal planes with R = 2.0, is optimal for motion-insensitive liver PDFF quantification.