Steady-state imaging methods, such as balanced steady-state free precession (bSSFP) are widely used in many applications, and in conjunction with high acceleration parallel imaging, this approach is highly valuable for dynamic or high-resolution imaging, for example, localized imaging of cranial nerves. Amplification of noise and incomplete aliasing artifact suppression are common issues with high acceleration rates. In this work, we explore the use of inner-volume imaging to improve image quality for highly accelerated MRI. In inner-volume imaging, a 2D or 3D sub-region within an object is excited such that only a reduced field-of-view (rFOV) needs to be encoded. 3D spatial selection is traditionally performed with a sequence of spin-echo RF pulses, however recent developments in gradient trajectory optimization and/or parallel transmit systems have enabled short 3D tailored pulses that directly excite the inner-volume [1]. Here, we propose to suppress aliasing from outer-volume signal by combining inner-volume steady-state imaging with parallel imaging (GRAPPA).We acquired two in vivo bSSFP brain data sets, one with conventional axial slab selection and one with an inner-volume excitation (IVex) RF pulse designed using the method in [1]. Experiments were run on a GE 3T scanner, all using the same full-FOV 3DFT readout (200x200x40 matrix; 24x24x20 cm FOV) with TR = 15ms and flip angle 15o. The IVex pulse was designed for a 6x6x3 cm^3 cube, without taking into account field map information. We performed 1-D GRAPPA along the in-plane phase-encode direction with 6 ACS lines and retrospective undersampling factors (outside the ACS lines) of 2, 4, 6, and 8 (net acceleration factors R= 1.75, 2.78, 3.45, and 3.64). Artifact levels within the inner-volume were compared against the ground truth full-FOV acquisition, for GRAPPA using both slab and IVex excitation pulses.Figures 1, 3 and 2, 4 show GRAPPA reconstruction results for slab and IVex excitations, respectively. In each of these images, the top row shows the reconstructed images from full k-space data and with GRAPPA reconstructions of different R, and the bottom row shows the corresponding error level relative to the ground truth (full k-space) image. Normalized root mean squared error inside the ROI is indicated in the top left corner in each error level image. With IVex, the residual image error inside the IV stays low and is greatly suppressed compared to the same region reconstructed with the conventional slab excitation pulse. A direct comparison of the absolute valued error between inner-volume excitation and slab excitation is shown in figure 5.

Imaging inner volumes rapidly can be accomplished using inner volume imaging or by using highly accelerated parallel imaging. For bSSFP, IVex alone is insufficient because imperfect excitation pulses still produce substantial outer volume signals, which aliases onto the image. Alternatively, highly accelerated parallel imaging could be used to image the full volume, but g-factor noise amplification and unsuppressed aliasing degrade image quality. In this work, we show that the combination of these approaches allow for reduced image acquisition times with limited artifacts. The improvement of the GRAPPA results arises from two related phenomena – the parallel imaging method, based on autocalibration, is better conditioned because there is less outer volume signal and further, the reduced outer volume signal reduces the corresponding aliasing. In many cases, bSSFP images are further distorted by off-resonance banding, so smaller volumes also offer the ability to shim and tune the acquisition of the volume of interest (VOI).

This general approach for 3D short TR imaging has been enabled by recent advances in multidimensional excitation and parallel excitation. We observe that the image contrast from IVEx is slightly different from conventional excitation sequence and care will need to be taken to ensure an adequate flip angle in the VOI.

We demonstrated that combined parallel imaging (GRAPPA) with 3D inner volume excitation can greatly reduce artifacts for inner-volume imaging with bSSFP.