## INTRODUCTION

Balanced steady-state free precession (SSFP) sequences (1–3) have gained popularity in magnetic resonance imaging (MRI) as they can yield high signal-to-noise ratios (SNR) within very short scan times. However, there are several problems limiting the applicability of SSFP imaging. The balanced SSFP signal is a function of the local resonant frequency, leading to characteristic signal nulls/voids (known as banding artifacts) in regions of large resonant frequency variation. Furthermore, the bright lipid signal is often undesired.

At higher field strengths or with longer repetition times (TR), the banding artifacts become more pronounced. It is therefore necessary to limit the off-resonance frequency variation to approximately 2/(3*TR) to avoid any banding artifacts (4). However, it is not always possible to limit the repetition time as specific absorption rate (SAR) considerations and resolution requirements may place constraints on the minimum TR. A longer minimum TR due to increased power deposition and resonant frequency variations at higher fields can potentially lead to severe banding artifacts.

A common strategy to reduce these artifacts has been to acquire a plurality of SSFP images, where the radio-frequency (RF) pulse phase increment between successive TRs is changed with each acquisition (5, 6) to shift the spectral response of the signal. Several methods for combining these multiple acquisitions have been proposed, including maximum-intensity (MI) (7), complex-sum (CS) (8), magnitude-sum and sum-of-squares (SOS) (9) combinations, and the nonlinear averaging reconstruction method (10). The complex-sum method aims to reduce banding artifacts but is far from optimal in terms of SNR efficiency. On the other hand, the magnitude-sum and SOS methods yield higher SNR efficiencies but provide less robust suppression of banding artifacts.

It is necessary to suppress the fat signal in applications where the tissue of interest has comparable or smaller signal than fat, including coronary artery imaging (11), cartilage imaging (12) and flow-independent angiography (13). There are various methods for suppression; one common way of reducing the fat signal in SSFP is to shape the periodic frequency response such that a broad range of frequencies around the resonant frequency of lipid are selectively masked out. Recent examples of this group of methods include fluctuating equilibrium magnetic resonance (FEMR) (14), linear combination SSFP (LC-SSFP) (8), binomial excitation patterns (15), periodic flip angle variations (16, 17), and fat suppressing alternating TR (FS-ATR) SSFP (18).

A drawback of these methods is the wedge shape of the stop-bands. The relatively broad stop-bands fail to yield suppression over certain ranges of frequencies, leading to a residual fat signal comparable to the water signal. Consequently, moderate to large resonant frequency variations will compromise the robustness of fat suppression.

In this work, we present a new method (weighted-combination SSFP or WC-SSFP) for combining a plurality of SSFP images with different RF phase increments for improved shaping of the SSFP profile and demonstrate its applicability to banding artifact reduction and fat-water separation. Our method approaches the SNR efficiency of the SOS method, while reducing the banding artifacts as effectively as the CS method by weighting each SSFP dataset by a power (greater than 1) of its magnitude. The exact value of the power is a control parameter which adjusts the trade-off between banding artifact reduction and SNR efficiency, giving greater flexibility for image optimization. The favorable SNR efficiency properties and robust banding artifact reduction coupled with this flexibility to tune for specific applications will allow higher field SSFP imaging, higher resolution or reduced SAR imaging over a greater range of TRs and with reduced banding artifact.

We further propose a new SSFP combination method for improved fat suppression. The LC-SSFP method produces a stop-band centered at the fat resonance by combining two phase-cycled SSFP acquisitions. The two combined magnetization profiles are out-of-phase in the vicinity of the fat resonance. Consequently, the two profiles are subtracted from each other. Since the magnitudes of the subtracted profiles are not the same for all frequencies, there is residual stop-band signal in the final image. The performance of the LC-SSFP method degrades at higher flip angles and when the tissue sample has a relatively low T1/T2 ratio. Weighting SSFP datasets by a negative power (between −1 and 0) of their magnitudes and combining them as in LC-SSFP achieves a drastic improvement in suppression robustness without affecting the pass-band. The range of flip angles and T1/T2 ratios for which LC-SSFP works robustly are expanded. The level of stop-band suppression can be adjusted through the power control parameter to meet application-specific needs. 2D and 3D fat- or water-suppressed SSFP imaging in the presence of large off-resonant frequency variations and at higher resolutions can be successfully accomplished with the proposed method.