Recently, myelin water fraction (MWF) was investigated using multi-echo GRE data^1,2. Generally, to ensure phase consistency among the echoes, multi-echo acquisitions use unipolar gradients. However, these unipolar gradient multi-echo sequences reduce acquisition efficiency and increase echo spacing. To overcome this, bipolar gradient can be used to improve the acquisition efficiency and reduce echo spacing. However, in bipolar gradient system, k-space misregistration induced by readout gradient delays and eddy-currents can induce phase errors³. Thus, myelin water imaging using bipolar gradients has not been proposed due to its severe artifacts. In this work, we present a MWF using bipolar gradient multi-echo GRE sequence with a k-space shift correction³.

[k-space shift correction] Gradient delays and eddy currents result in delays of echo acquisition. These make k-space misregistration and lead to phase discrepancy between odd echoes and even echoes. To compensate for the k-space misregistration, all echoes (even and odd echoes) are used to estimate the k-space shift distance. The detailed procedure is as follows: 1. Signals from each echo is 1D Fourier transformed. 2. Calculate cross-correlation between the odd and even echoes 3. K-space shift is estimated from the slope of the cross-correlation signal’s phase using linear regression. 4. The shift distance is obtained from the estimated phase slope by shift = slope*base-resolution/(2*pi). The shift distance is calculated for each channel, and then averaged.

[Data acquisition and processing] A healthy volunteer was scanned using both unipolar and bipolar sequences at 3T (Tim Trio, Siemens Medical Solutions, Erlangen, Germany) with a 4 channel head coil. For MWI, multi-echo 3D GRE sequence was used. The imaging parameters were as follows: matrix size: 128x128x32, spatial resolution 2x2x3mm³, TR = 84ms, TE₁ = 1.65 ms, ΔTE = 1.04 ms for bipolar gradient and 2.08 ms for unipolar gradient, # of echoes = 30 for bipolar gradient and 16 for unipolar gradient, flip angle = 30°, BW = 1560 Hz/Px. The total scan time was 5 min 44 sec. After k-space shift correction, a three-pool complex model which includes frequency offset terms (Δf) was fitted to each voxel to estimate MWF².

Fig.1 shows the result of the k-space shift correction. Fig.1 (top) plots the magnitude of the first two k-space echo data projected along the readout axis. The k-space misalignment introduces a linear phase discrepancy between even echoes and odd echoes as shown in Fig.1 (bottom). After correction, this discrepancy is corrected. Also, Fig.2 shows the phase at a chosen voxel (yellow arrow) with respect to echo time in bipolar gradient sequence. The phase was proportional to the echo time after correction. Fig.3 shows the MWF with 30 bipolar echoes fitted with and without correction of the k-space shift. Without the k-space shift correction, the fitting failed to obtain correct MWF and Δfmy-ex map. Fig.4 shows the result of MWF with unipolar and bipolar gradient GRE data. When the bipolar multi-echo GRE is used, quantification and SNR is improved.In this study, we demonstrated that MWF with bipolar gradient can be performed by correcting k-space shift. The improved scan efficiency of the bipolar sequence may translate into a reduction in the scanning time. Compared to unipolar MWF, bipolar MWF yields a reduction in ΔTE, which leads to improved SNR and more accurate quantification.