Combined with the use of parallel imaging, single shot EPI is the most common choice for diffusion weighted imaging and functional MRI studies. A per-volume or even per-slice basis reference scan is usually needed for correcting the phase inconsistency between odd and even shots attributed to eddy current. Not only the reference scan is time consuming compared to the EPI acquisition itself, the phase variation may be temporally varying that calls for dynamic reference scan. Furthermore, the reference scan is limited to first order correction, whereas EPI acquisitions are often confounded with high order phase contamination in cases such as oblique scans. In this work, we extend the concept of multiplexed SENSE [1,2] and further incorporate the between shot phase inconsistency as part of coil sensitivity weighting and to eliminate such inconsistency simultaneously with the SENSE correction.

The proposed method is illustrated in flow chart in Fig.1. Firstly, for an R-acceleration factor SENSE acquisition (R could be 1), the acquired k-space is split into two parts, namely the even k-space lines and the odd k-space lines. Conventional SENSE reconstruction is first performed separately on the two k-spaces leading to two reconstructed N/2 ghosted images. Then the residual phases in the two reconstructed images are collected and treated as the phase error attributed to eddy currents. A low-pass filtering is then performed on the two phase maps to remove noises or phase attributed to tissue susceptibility. The obtained phase maps are then used to update the original coil sensitivity map to form a set of composite coil sensitivity maps that contains N/2 phase errors. Notice at this point, the reconstruction matrix is doubled (Equ.(2), compared to regular SENSE recon in Equ.(1)) as we are dealing with twice as many virtual coils, however given the effective acceleration factor R the resulting g-factor penalty remains the same. In this way, the phase variations caused by eddy currents are intrinsically eliminated in the reconstructed image. In addition, this multiplexed SNESE phase correction may deal with higher order phase variations as determined by the form of the phase map filtering performed.

[MxR] = [MxR] [Rx1] (1)

[2MxR] = [2MxR] [Rx1] (2)

(1) regular SENSE recon (2) multiplexed SENSE recon

M: number of receiver coils R : SENSE acceleration factor

Both phantom and in-vivo experiments were performed to verify the performance of the proposed method using a whole body 3T scanner (GE Discovery 750) equipped with an 8-cahnnel brain coil. A uniform sphere phantom (for enhanced visibility of ghosting) and healthy volunteer (with consent obtained) were scanned with a single shot SE-EPI and DW-EPI (b=600) sequence with a SENSE acceleration factor of 2. Acquisition matrix size used was 128x128 and FOV was 22cm. Oblique plane scan was performed. Standard reference scan corrections and the proposed multiplexed SENSE corrections were performed to correct for the N/2 ghost on the same data set retrospectively. Fig. 2 compares the N/2 ghost corrections using standard reference scan and the proposed multiplexed SENSE method. It is seen that in the phantom results, the residual ghosting is very obvious using the standard reference scan correction due to eddy current contamination at oblique plane. It is also noticed that the ghosting was 4 fold despite a SENSE factor of 2, as the N/2 ghost is further multiplexed with SENSE aliasing. The ghosting is much reduced in the images obtained using the multiplexed SENSE. A stronger level of residual ghosting is also seen with the SE-EPI in-vivo data with standard correction, whereas the residual ghosting is much less visible in the DW-EPI acquisition as the diffusion gradient helps to weaken the ghosting signal. It is also seen that there is no additional SNR penalty using the proposed method as expected. In this work, the N/2 ghost in EPI is corrected based on multiplexed SNESE that treats the phase variations between the even-odd lines as part of the coil sensitivity. Oblique plane phantom and in-vivo experiment showed much reduced residual ghosting level with the proposed method as compared to standard reference scan, which is potentially due to the high order phase variations that are uncorrected in the latter case. This method may also be beneficial for multi-phase acquisitions such as fMRI where the phase inconsistency could be temporally varying.