Blipped controlled aliasing parallel (blipped-CAIPI) imaging is widely used for fast imaging, which is one of the simultaneous multi-slice (SMS) imaging methods^[1]. Water-fat separation methods such as DIXON and iterative decomposition of water and fat with echo asymmetry and least-squares estimation^[2] can be combined with the blipped-CAIPI technique. However, it results in the chemical-shift ghost artifact because fat signal on slightly shifted position is exited in the slice-selection process. This geometric error in slice-selection generates additional phase cycling, which causes the ghost artifacts on each slice’s fat image. In this abstract, a sensitivity encoding (SENSE)-based water-fat separation method is proposed, which considers the additional phase cycling on fat signal and obtains more accurate water-fat separated images.

A slice-selection process excites fat signal on shifted position from the region of water signal due to chemical shift. The geometrical slice-selection error can be calculated as follows.

\[z_{shift} = \frac{f_{fat}}{\gamma G_{ss}}\tag1\]

where z_shift is the distance of fat shift from the water slice, f_fat is the chemical shift frequency between fat and water, γ is the gyromagnetic ratio, and G_ss is a slice-selection gradient. This geometric error generates additional phase cycling of fat signal by gradient blip of the blipped-CAIPI (Figure 1). This additional phase cycling makes the ghost artifact on fat images in phase-encoding direction. The additional phase is calculated as follows.

\[\alpha = 2\pi \gamma M_{blip}z_{shift}=\frac{2\pi f_{fat}M_{blip}}{G_{ss}}\tag2\]

where α is an additional phase, and M_blip is the moment by blip gradient.

Figure 2 shows the phase variation in k-space and its point spread function (PSF) for two-slices blipped-CAIPI. Unwanted peak in fat signal is generated on N/2-field of view (FOV) shifted position. Therefore, the ghost of first slice’s fat signal is overlaid on the second slice’s fat signal. The ghost artifacts cannot be separated by using conventional water-fat separation methods. However, the additional phase α can be calculated from equation (2), from which the ghost artifacts can be easily decomposed by SENSE-based least square method.

\[\rho = \underset{\rho}{\operatorname{argmin}}{\parallel S-C\cdot \Gamma \cdot B \cdot A\cdot F_u \cdot \rho \parallel}\tag3\]

where ρ is the reconstructed signal intensity, S is the acquired signal, C is a coil sensitivity, Γ is the additional phase, B is the blipped phase for blipped-CAIPI, A is chemical shift, and F_u is the under-sampled Fourier transform.

Figure 3 shows a simulation results representing the chemical shift effect in the blipped-CAIPI. This simulation is based on the blipped-CAIPI echo planar imaging (EPI) sequence for two-slices SMS imaging. Water and fat images are separated using the fat shift in phase-encoding direction caused by the phase variation between water and fat, which is generated during echo train. Because the distance of fat shift can be calculated, water-fat separated images are well reconstructed by using the conventional parallel imaging, which is referred as EPI chemical shift separation^[3]. However, the conventional -unknown EPI chemical shift separation method cannot reconstruct the accurate water and fat images from the acquired image due to chemical shift ghost artifact. The proposed method reconstructs each image correctly.

Experiments were conducted on a 3.0 T MRI scanner (Siemens Verio, Germany) and 32-channel head coil. Blipped-CAIPI multi-shot fly-back EPI sequence was implemented for two-slices SMS imaging with the imaging parameters of 4000_ms/25_ms/1560_Hz/Px(TR/TE_effective/BW_readout), image size of 64×128, voxel size of 2.5×2.5×5.0_mm³, the distance between slices of 20_mm, and G_ss of 4.95_mT/m, which means α is 18.52° .

Figure 4 shows experimental results. For comparison, multi-shot fly-back EPI sequence is used, and images are reconstructed by EPI chemical shift separation method, which uses the phase variation during echo train and coil sensitivity. Without considering additional phase cycling, the chemical-shift ghost artifact in fat images is shown as expected. The proposed reconstruction method separates water and fat images correctly with the calculated α of 18.52°.Simulation and phantom experiment show that there is the chemical-shift ghost artifact in fat images from the conventional water-fat separation method. This artifact is generated by additional phase cycling of fat signal. However, the additional phase cycling generated by chemical shift can be calculated, and the correct water-fat separated images can be acquired by the proposed SENSE-based reconstruction method.