The g-ratio of a nerve fiber, defined as the ratio of axonal diameter (excluding myelin thickness) to fiber diameter (including myelin thickness), is an important biophysical parameter that determines conduction velocity of the fiber [1]. Recently, an MRI method that measures an aggregated g-ratio of a voxel has been proposed [2] using quantitative magnetization transfer (qMT) imaging [3] and neurite orientation dispersion and density imaging (NODDI) [4] to estimate the parameters for the g-ratio formula: myelin volume fraction (MVF) from qMT and axon volume fraction (AVF) from NODDI. However, the qMT imaging requires a substantial scan time (over 30 min) [3] and the resulting fractional pool size (F) requires a scaling coefficient which may need complex histological investigation to translate it to MVF [2]. Recently, GRE myelin water imaging (GRE-MWI) that utilizes multi-echo GRE data to estimate a fast decaying myelin water signal has been proposed as an approach to acquire myelin water fraction (MWF) in a relatively short scan time (<10 min) [5-6]. The resulting MWF is a quantitative measure of myelin water volume and can be converted to MVF using available information (described below). Hence combining GRE-MWI and NODDI can reduce the total scan time and may improve quantification for the g-ratio imaging. In this study, we demonstrate that GRE-MWI and NODDI can deliver a high quality g-ratio map of the whole brain in a reasonable scan time.[Data acquisition] Multi-echo GRE data for GRE-MWI and multi-shell diffusion imaging data for NODDI were acquired in four healthy subjects (IRB approved) at 3T. Scan parameters for 3D multi-echo GRE were: TR=67ms, TE=1.52:2.03:31.97ms, flip angle=30°, FOV=256×256×100mm³, resolution=2×2×2mm³, number of echoes=16, and scan time=7min. For NODDI, a three-shell diffusion imaging (b=300s/mm² with 8 directions; b=700s/mm² with 32 directions; b=2000s/mm2 with 64 directions; b=0s/mm² with 13 averages) was acquired using a multi-band (band=2) SE-EPI diffusion sequence. The same resolution, slice thickness and slice number as the GRE sequence were used with TR/TE=4000/95ms, and scan time=9.75min. [Data processing] From the multi-echo GRE data, the signal intensities of myelin water and axonal/interstitial water were estimated by using the GRE-MWI algorithm described in [6]. The resulting signal fraction was translated into the volume fraction with the following considerations. Myelin sheath contains alternating layers of lipid (60% of total volume) and water (40% of total volume) [7-8] and, therefore, the myelin water signal needs to be scaled by 1/0.4 to represent the total myelin volume. The proton density of axon and interstitial space is approximately 0.85 [9] and, therefore, the axonal/interstitial water signal needs to be scaled by 0.85 to represent the axonal/interstitial water volume. As a result, the MVF can be estimated by ([myelin water signal]/0.4)/([axon and interstitial water signal]/0.85+[myelin water signal]/0.4). For AVF estimation, the fraction of free, restricted, and hindered diffusion compartments was calculated using NODDI. Among them, only the restricted diffusion compartment contributes to AVF. Since myelin signal was not detected in NODDI because of long TE of DTI, the restricted diffusion compartment $$$((1-\nu_{iso})\nu_{ic})$$$ did not contain MVF. Hence, the restricted compartment was scaled by (1-MVF) and AVF became $$$((1-MVF)(1-\nu_{iso})\nu_{ic})$$$ [10]. Finally, the aggregated g-ratio was alculated by $$$g=\sqrt{1/(1+MVF/AVF)}$$$ [2]. [Data analysis] ROIs were chosen in optic radiation (OR), corticospinal tract (CST), superior longitudinal fasciculus (SLF), and splenium (SPL) to calculate g-ratios and to test the consistency of our method across the subjects. The MVF, AVF, and g-ratio maps are shown in Figure 1. The subject averaged ROI values are listed in Table 1. A high g-ratio was observed in CST because of low MVF and high AVF whereas a low g-ratio was found in OR because of high MVF and low AVF. These results are partially supported by histological evidences that CST has large axons [11-12] whereas OR is populated with thick myelin sheath [13]. The mean g-ratios of the ROIs ranged from 0.68 to 0.86. These values are close to an optimum g-ratio of 0.77 for signal conduction [14]. Figure 2 demonstrates consistent g-ratio values across the subjects.In this work we demonstrated an approach of generating a g-ratio map using GRE-MWI and NODDI. The scan time was less than 20 min for the whole brain coverage in a 2 mm isotropic voxel size. Further reduction in the DTI scan time is possible with a powerful gradient system [15]. Compared to a recent result using qMT and NODDI, which revealed a relatively uniform g-ratio map [2], our map shows larger g-ratio variations among fiber bundles. This may be in line with the histological measurements showing variations in g-ratio [14,16].