Quantitative susceptibility mapping (QSM) provides direct measurements of the molecular composition and cellular architecture of the cerebral tissue (1, 2). Because the oxygenation saturation of blood affects its magnetic susceptibility, QSM has been recently compared against conventional magnitude based BOLD fMRI for detecting neuronal function under visual or motor stimulation (3, 4). It has also been investigated for measuring cerebral metabolic rate of oxygen and oxygen extraction fraction (5). To date, the most widely used acquisition method for QSM is gradient echo (GE) imaging. However, its relatively lengthy acquisition time limits its clinical application in motion-prone situations, investigation of high frequency functional activities and real-time measurement of cerebral physiology. Simultaneous multi-band imaging can be used to accelerate QSM based on 2D GE-EPI, by increasing the sampling rate thus shorten acquisition time. Here we present preliminary results of a simultaneous multi-band accelerated 2D EPI acquisition for QSM and a correction method for phase gap between bands.

Pulse Sequence

In our preliminary implementation, we acquired a dataset using a 2-band and 2-shot interleaved EPI sequence with a Nova 32-channel phase array RF coil on a 3T scanner. Imaging parameters were: in-plane (sagittal plane) matrix size = 128 *128, FOV = 25.6 * 25.6 cm2, thickness = 3 mm, TR = 2 sec, TE = 25 msec. Magnitude and phase images were reconstructed using the generalized MUSE algorithm to effectively eliminate both in-plane and through-plane aliasing artifacts (6). Figure 1 shows original phase of multi-band acquisition without correction.

Phase Correction

In multi-band phase, larger number of wraps was identified in band2 as compared to band1 in axial plane. Discontinuity of phase between band1 and band2 was also observed in coronal plane. This is very likely resulted from misalignment of k-space across these two bands, which can be corrected by adding a linear phase term to band2. Corresponding equation is:

S_band1 = S_band2 × exp⁡[i × (f1 × x+f2 × y+f3)] Equation 1

To eliminate phase gaps between bands, number of wraps of band2 should be the same as that of band1 in axial plane. Number of wraps can be represented by sum of phase gradient along x direction (figure 2(a)), which is denoted as cost function of number of wraps, cost_nw. Phase continuity in axial and coronal plane can be reflect by sum of phase gradient along z direction in the center adjacent two slices, which are denoted as cost functions of continuity in axial plane (cost_ca, figure 2(b)) and coronal plane (cost_cc, figure 2(c)) respectively. Total cost function can then be written as:

cost= λ × cost_nw (band1) / cost_nw (band2)+ cost_ca + cost_cc

where λ is a weighting factor. In equation 1, f1, f2, f3 can be estimated using unconstrained nonlinear optimization through minimizing the above total cost function.

QSM reconstruction

Multi-band phase images corrected using the proposed method were further processed using 3D Laplacian phase unwrap, 3D VSHARP background phase removal (7) and STAR dipole inversion (8). As a comparison, multi-band phase images were further processed using 2D Laplacian phase unwrap across sagittal plane, followed by 2D VSHARP background phase removal and STAR dipole inversion (8).

Figure 3 illustrates corrected phase using the proposed method. Phase of band1 is nearly perfectly aligned with phase of band2 in both axial and coronal planes. Figure 4 illustrated quantitative susceptibility maps derived from uncorrected multi-band phase, corrected multi-band phase and 2D processed multi-band phase. QSM map of uncorrected data showed substantially amplified error caused be phase gap between the two bands. Significant wave-pattern artifacts can be identified in the center 4 slices in uncorrected data. The proposed method markedly improved the image quality with nearly no artifact observable in the center slices and meanwhile retained the susceptibility contrast especially across deep gray matter nuclei and surrounding tissues. Although 2D processing can overcome phase gap induced artifact in center slices, significantly reduced susceptibility value and decreased contrast in the deep gray matter regions can be visualized. Besides, additional artifacts in cortical regions were induced by incomplete background phase removal. We demonstrated the feasibility of using the proposed phase correction method for increasing the accuracy of QSM reconstruction from multi-band acquisitions. With multi-band acquisition, we were able to greatly shorten data acquisition time. It is expected that facilitate this method would benefit further clinical application of QSM and QSM based cerebral functional and physiological studies.