Balanced steady-state free precession (bSSFP) is commonly used for cardiac cine imaging due to its high SNR and desirable myocardium-blood contrast, but suffers from signal nulls due to off-resonance. Acquiring multiple images with different RF phase-cycling enables the reconstruction of a combined image with reduced banding, but extra stabilization periods are needed between phase-cycles, and the sequential acquisition of different phase-cycles may cause registration errors during combination. Foxall ¹ proposed slow frequency modulation to obviate stabilization periods between phase-cycles. By modifying the slow frequency modulation scheme for use in the heart, we developed a bSSFP cine sequence that can acquire interleaved phase-cycles within a breath-hold.

To test the feasibility of slow frequency modulation for cardiac cine imaging, a sequence with two phase-cycles was developed. Because a constant heart rate cannot be assumed, a modified scheme (Fig. 1) was designed in which the phase-cycling modulates slowly for a time equal to the shortest expected RR interval, after which the phase increment remains constant until the next cardiac trigger. Before the scan, we determine the number of TRs, n_FM, that covers the shortest expected RR interval. For n_FM TRs, the phase increment increases by 180^o/n_FM each TR so that the other phase-cycle is reached before the next trigger. This synchronization of the phase increments with the cardiac triggers offers immunity to heart-rate variability – a particular cardiac phase will be acquired with the same average phase-cycling during even heartbeats, and the average phase-cycling during odd heartbeats will be offset by exactly 180^o.

A discrete phase-cycled sequence was also implemented for comparison. Both sequences are prospectively triggered and begin with two heartbeats of dummy acquisitions. For the slow frequency modulation sequence, interleaved phase-cycled images are acquired over 26 heartbeats. For the discrete phase-cycled sequence, a 90^o RF phase-cycled image is acquired (13 heartbeats), followed by a stabilization period (2 heartbeats), and then a 270^o phase-cycled image is acquired (13 heartbeats). Both sequences are sequential, segmented 2DFT acquisitions with 10 views per segment (41-ms temporal resolution). Each k-space segment is repeated until the next trigger, and, within each segment, an odd-even view ordering is used to cause temporal artifacts to ghost outside of the heart (e.g., the first segment’s ordering was 1-3-5-7-9-10-8-6-4-2).

To verify that the constant phase-cycling at the end of each heartbeat does not disrupt the slow frequency modulation profile, we used a phantom doped with nickel chloride and manganese chloride to have blood-like T1/T2 = 1045 ms/207 ms. We examined the spectral profile during the slow-frequency modulation TRs that follow the constant phase-increment TRs by acquiring projections of this phantom with no phase-encoding. A heart rate varying between 60 and 73 bpm (corresponding to 992 ms and 820 ms RR intervals, respectively) was simulated.

Slow-frequency-modulated and discrete phase-cycled scans of two volunteers were performed on a 1.5T GE Signa Excite scanner using an 8-channel cardiac receive coil and the following parameters: 22x22 cm² FOV, 1.15x1.7 mm² resolution, 8 mm slice thickness, 60° flip angle, 1.8 ms TE, 4.1 ms TR, and 28 heartbeat scan duration for slow frequency modulation and 30 heartbeat scan duration for discrete phase-cycling. To simulate increased off-resonance, the Y linear gradient was deliberately offset from the optimal shim, or “de-shimmed.” Phantom data to examine the spectral profile was acquired with a quadrature transmit/receive head coil.

Phantom studies show that, over a significant range in RR interval length, the slow-frequency-modulation profile experiences very limited transients from the constant phase-increment TRs at the end of each simulated heartbeat (Fig. 2). Thus, the proposed phase-cycling scheme seems appropriate for using slow frequency modulation with variable heart rates.

The bands in the two phase-cycled images in Fig. 3 are offset, facilitating the reconstruction of a combined image without signal nulls. The root-sum-of-squares-combined images from the proposed slow-frequency-modulated method (bottom row of Fig. 4) are comparable to those from standard phase-cycling (top row) but were acquired in a shorter scan time since frequency modulation eliminates the stabilization periods between phase cycles. Both methods greatly reduce the dark banding artifacts.

Preliminary in-vivo results show the feasibility of cardiac cine imaging using slow frequency modulation within a breath-hold for banding-artifact reduction. Slow frequency modulation leads to interleaved acquisition of the phase-cycles and a shorter scan time than discrete phase-cycling. Acceleration via parallel imaging and partial Fourier should shorten the scan duration to less than a standard breath-hold. Slow frequency modulation, combined with this acceleration, would enable the acquisition of more phase-cycles within a breath-hold, which would facilitate a flatter spectral profile after phase-cycle combination.