Balanced steady state free precession (SSFP, also known as True-FISP, FIESTA, or Balanced-FFE) has emerged as a powerful technique for rapid magnetic resonance imaging. Compared to gradient-echo techniques, SSFP provides superior SNR efficiency and image quality, but is highly sensitive to resonance frequency (1–4). In particular, the usable bandwidth of conventional SSFP is less than 1/TR. Signal nulls occur every 1/TR in resonance frequency, producing well-known “banding” artifacts in images. More global transient artifacts occur when bands fall in regions of flow (5).
In static tissue, multiple-acquisition methods using phase-cycling (4, 6, 7) successfully eliminate banding, but require multiple steady-states to be reached, and increase scan-time by at least a factor of two. Banding can also be avoided by shortening the TR (8). However, this reduces the maximum readout duration and limits spatial resolution. Shortening the TR also reduces the readout duty-cycle, and hence SNR efficiency, and may prevent the use of time-efficient k-space trajectories such as echo-planar (9, 10) or spiral (11).
The hypothesis of this work is that alternating repetition times can be used to widen band-spacing of SSFP beyond 1/TR, and therefore reduce and potentially avoid banding artifacts in single-acquisition SSFP MRI. This technique, called wideband SSFP, uses two alternating repetition times (12) with alternating RF phase to establish a band spacing that is up to two times wider than that of conventional SSFP, with a modest increase in scan-time (13).
We first present a theoretical description of wideband SSFP and the mechanism for increasing band spacing, and analytic approximations for the resulting image contrast and SNR efficiency compared to conventional SSFP. Simulations are used to accurately determine the image contrast, optimal flip angle, and SNR efficiency of wideband SSFP. Wideband SSFP signal profiles are verified in a phantom study. Finally, the proposed method is applied in vivo to banding artifact reduction in cine ventricular function imaging and high-resolution cartilage imaging at 3 Tesla.
The SSFP pulse sequence consists of a rapidly repeating sequence of fully balanced excitations and acquisitions. All gradients have zero net area over the course of one repetition (∫G(t)dt = 0), and should also have a zero gradient first-moment when imaging in areas of flow (∫G(t)dt = 0) (11). Compared to gradient echo techniques, balanced SSFP provides superior SNR efficiency and strong T2/T1 contrast. For these reasons, balanced SSFP has proved useful in many cardiovascular, neurologic, body, and musculoskeletal applications.
One of two primary drawbacks of SSFP (the other one is RF heating) is its sensitivity to resonance frequency (1–4), with signal nulls occurring every 1/TR. These signal nulls often obscure anatomical structures, cause global artifacts due to transient oscillations and flow (5, 14, 15) and reduce the dianostic image quality.
Alternating the repetition time (TR) between two values with appropriate phase-cycling, has been successfully used to modify the spectral response of SSFP, and has been demonstrated as a means for achieving fat suppression (12, 16). The aim of this work is to use alternating-TR to widen the band spacing of SSFP and minimize undesired signal nulling (13). This technique is referred to as wideband SSFP, and is illustrated in Fig. 1. Two alternating repetition times (TRs and TR) with alternating RF phase are used to produce an oscillating steady state with two distinct echoes (black and white circles). The primary design parameter is the ratio of the short and long repetition times, which we define as a = TRs/TR, where 0 < a ≤ 1. Both echoes may be used for imaging, however, as a becomes small, it may be impractical to acquire data during TRs (white circle). Note that when a = 1, this sequence is equivalent to conventional SSFP.
Mechanism for Increasing the Band Spacing
Figure 2 illustrates the steady-state magnetization path and frequency response of wideband SSFP with a = 1.0, 0.7, and 0.4. Five isochromats are shown, exhibiting different amounts of precession during TR: 0.1π, 0.3π, 0.5π, 0.7π, and 0.9π. In conventional SSFP (a = 1.0, Fig. 2a–c), precession and relaxation are the same in both TR intervals, resulting in a symmetric steady-state. As the precession over a TR approaches π, the steady-state magnetization and signal approaches zero, which produces the well-known “banding” artifact in SSFP MRI (1, 3) at resonance frequencies , , and periodically thereafter.
Wideband SSFP tilts the magnetization path to one side (see Fig. 2d–i), permitting a greater amount of precession during TR before signal nulling occurs. The precession over TR can be greater than π and the net deflection (along Mx in Fig. 2) will still be balanced by correspondingly smaller amounts of precession during TRs. The echo during TR (black circle, solid black profile) has a wide flat passband with all depicted isochromats exhibiting nearly identical transverse magnetization, therefore producing uniform signal. The echo during TRs (white circle, dashed green profile) has higher and less-uniform signal across the passband.
Band Spacing of Wideband SSFP
The null-to-null spacing as a function of a is shown in Fig. 3. The band spacing is not a function of T2/T1, but does depend on flip angle, α. As α is reduced, the band spacing approaches or equivalently, . This empirical approximation is used for the remainder of this manuscript and has an error of less than 5% for α < 60°, where error is measured as the difference between the approximated band spacing and true band spacing divided by the true band spacing.
The ability of this method to modify the SSFP frequency response is illustrated in Fig. 4. For a fixed imaging TR, wideband SSFP can be used to widen the band spacing (see Fig. 4a). As TRs is shortened (with TR fixed), there is a decrease in signal intensity and a widening of the passband. In this way, wideband SSFP may be used to reduce and potentially avoid banding artifacts. For a fixed band spacing, wideband SSFP can be used to increase the available readout duration (see Fig. 4b) As TR is increased (with TR + TRs fixed), there is a decrease in signal intensity but the passband width is maintained. In this way, wideband SSFP may be used to increase the available readout duration and (1) achieve high spatial resolution, which is limited by the amount of gradient area that will fit within TR, or (2) use time-efficient readout schemes (9–11).
The steady state magnetization at the imaging echo can be derived analytically from the Bloch equations in matrix form (17). On-resonance, the magnitude of the transverse magnetization in the steady state simplifies to:
Note that when a = 1.0 (conventional SSFP), this is equivalent to Eq.  from Ref.18. When α = 90°, Eq.  simpli- fies to:
This suggests that wideband SSFP exhibits T2/T1-like contrast similar to that of conventional balanced SSFP.
Numerical Bloch simulation can be used to determine the contrast over the relevant range of resonance offsets. Figure 5 illustrates the steady-state passband signal as a function of T2/T1, for different a-values and approximately the same band spacing (i.e. with TR + TRs fixed). The plotted signal amplitude is relative to Mo, and is an average over 2/3 of the null-to-null spacing in the spectral profile of each sequence. Smaller a-values generate weaker signal and contrast during TR and stronger signal and contrast during TRs.
Optimal Flip Angle for Wideband SSFP
The choice of imaging flip angle is typically made to maximize SNR efficiency while operating within SAR limits. The steady-state signal is a function of both a and α, and therefore the SNR or CNR optimal flip angle will vary with α. Figure 6 illustrates the myocardial (T1/T2 = 1100/40 msec) and blood (T1/T2 = 1500/140 msec) signal at 3T for different flip angles and a-values. Notice that the optimal flip angle for wideband SSFP is comparable to that of conventional SSFP when a > 0.3.
Central Signal Dip
The signal profile of wideband SSFP exhibits an intrinsic signal dip close to on-resonance (Δ f = 0) when relaxation and other phase effects become significant relative to precession (see Fig. 7). Our empirical observations are that the size of the dip is a function of T2/T1 (deeper for lower ratios), and the width of the dip is a function of absolute T2/TR and T1/TR (widens for lower values). For the 3T cardiac study parameters described below, the signal dip size is 3.6% and 4.1% for myocardium and blood, respectively. Although small, such a dip may appear as a narrow signal band in tissues exactly on-resonance, and may cause mild transient artifacts if there is flow or motion through this region.
The SNR efficiency of wideband SSFP depends heavily on whether one or both echoes are used, and on the duration of excitation and other pulses that influence the available readout duration. As the a-value is reduced, the signal during TR decreases (see Eq. ) and the signal during TRs increases. Here, we consider the relative SNR efficiency of wideband SSFP in the case where TRs is not used for data collection.
Considering the conservative case where only TR is used for imaging, and comparing wideband SSFP with conventional SSFP with the same TR: (1) the scan-time is (1 + a) times longer, and (2) the passband signal is lowered by a factor, called S. As such, the resulting SNR efficiency would be:
where μw and μc represent the SNR efficiency of wideband and conventional SSFP, respectively, and μ is defined as , Δ v is the voxel size, and Tscan is the total scan time. Note that S is always less than 1, even when using the optimal flip angle for wideband SSFP (see Fig. 6). The SNR efficiency of wideband SSFP (when used to avoid banding artifacts) is always lower than that of conventional SSFP. In static imaging scenarios, where multiple-NEX are needed to achieve adequate SNR, multiple-acquisition methods will remain the method of choice for removing banding artifacts (4, 6, 7), because of their SNR efficiency. In cases where multiple acquisition methods are not applicable, or where SNR can be traded for reduced banding (common in high-field SSFP imaging), wideband SSFP will be faster than multiple-acquisition methods by a factor of at least 2/(1 + a).
Considering the case where only TR is used for imaging, and comparing wideband SSFP with conventional SSFP with the same approximate band spacing (i.e. , where TRc is the repetition time of conventional SSFP): (1) the scan-time is two times longer, (2) the pass-band signal is lowered by a factor S, and (3) the available readout duration is increased by a factor x. The resulting SNR efficiency would be:
by defining Td as the amount of time per TR that is not usable for data acquisition, x can be expressed as:
where TR and TRc are the repetition times for wideband and conventional SSFP, respectively. In this case, the SNR efficiency of wideband SSFP will be superior to that of conventional SSFP when Td is a substantial fraction of TRc, therefore making x large. This can be the case in high-field imaging applications where the TRc is limited to a few milliseconds because of off-resonance effects within a region of interest.
Accelerating the approach to steady state is important when combining this technique with any type of contrast preparation (e.g. fat saturation or inversion recovery). The single “α/2” tip approach (19) can be easily adapted to align the direction of with the steady-state magnetization at the echo during either TR or TRs. The excitation angle should be the angle between the steady-state magnetization at the echo and the longitudinal axis, which for wideband SSFP, is a function of both α and a. Figure 8 illustrates the simulated reduction in transient oscillations for myocardium at 3T, when using a single tip preparatory pulse to tilt to the axis of during TR.
MATERIALS AND METHODS
Experiments were performed on a Signa Excite HD 3T scanner (GE Healthcare, Waukesha, WI) with gradients capable of 40 mT/m amplitude and 150 mT/m/msec slew rate, and receiver capable of ±250 kHz bandwidth (2 μsec sampling). In phantom studies, a transmit/receive birdcage head coil was used. In cardiac studies, the body coil was used for transmission, and an 8-channel cardiac phased array was used for signal reception. In knee studies, the body coil was used for transmission and a 5-inch surface coil was used for signal reception. Scan volunteers provided written informed consent, and the imaging protocol was approved by our institutional review board.
A phantom study was performed to measure the spectral response and band-spacing of wideband SSFP. A uniform ball phantom (T1/T2 ≈ 150/30 msec) was imaged with centered 2DFT readouts during both echoes. A linear shim was applied along the frequency-encoding direction to create in-plane off-resonance. The scan parameters were flip angle = 30°, TR = 6 msec, and a = 1.0, 0.75, and 0.5. The measured spectral response was compared with Bloch simulations performed using MATLAB (Mathworks, South Natick, MA). Simulations used the actual gradient and RF waveforms from the experiments, and utilized the hard-pulse approximation with 4 μsec timesteps (17).
Wideband SSFP was applied to breath-held gated CINE cardiac imaging in three healthy volunteers. Localized center-frequency adjustment was performed over the left ventricle, and short-axis scan-planes were prescribed. Imaging parameters were: 2DFT acquisition, FOV = 30 × 30 cm2, resolution = 1.2 × 1.2 × 8 mm3, α = 30°, TR = 4.4 msec, plethysmograph gating, 16 R-R breath-hold. Data was acquired using conventional SSFP (a = 1) resulting in a temporal resolution of 76.8 msec, and with wideband SSFP (a = 0.45) resulting in a temporal resolution of 115.2 msec. In the case of wideband SSFP, no data was collected during TRs resulting in the lower temporal resolution. The expected band spacing for conventional and wideband acquisitions was 227 Hz and 313 Hz, respectively.
Wideband SSFP was applied to slice-selective knee imaging in two healthy volunteers. Imaging parameters were: 2DFT acquisition, FOV = 15 × 15 cm2, low resolution = 0.3 × 0.3 × 5 mm3, high resolution = 0.15 × 0.3 × 5 mm3. Data was acquired using low resolution conventional SSFP (a = 1, α = 25°, 10 sec scan time), high resolution conventional SSFP (a = 1, α = 22°, 22 sec scan time), and high resolution wideband SSFP (a = 0.26, α = 31°, 35 sec scan time). The flip angle choice was optimized for cartilage. In the case of wideband SSFP, no data was collected during TRs.
Figure 9 contains images and signal profiles from the phantom experiment. The band spacing in simulated and measured spectral responses showed excellent agreement. The measured null-to-null spacing of wideband SSFP with a = 0.75 and a = 0.5 were 11 and 34% larger than conventional SSFP (a = 1.0). As expected, the signal intensity of the long TR echo was observed to be lower than that of the short TR echo.
Figure 10 contains 3T cardiac images from one representative study. Conventional SSFP images exhibited banding artifacts (white arrow) and transient artifacts. When viewing these data in a cine loop, it is clear that the transient artifacts are due to flow. In all studies, wideband SSFP successfully widened the band spacing, shifted the banding artifact outside of the heart, and reduced flow-related transient artifacts, presumably due to the uniformity of the wideband SSFP signal profile. As expected, the wideband SSFP images show slightly reduced blood-myocardium contrast compared to conventional SSFP.
Figure 11 contains 3T knee images from one representative study. Low resolution SSFP images were free of banding artifacts, while high-resolution SSFP experienced banding artifacts (white arrows) due to the required increase in readout length and increase in TR. Wideband SSFP successfully suppressed the banding artifact, and exhibited the expected high spatial resolution compared to low-resolution SSFP with the same band spacing.
We have demonstrated that alternating repetition times can be used to widen the band spacing in the frequency response of SSFP. For a fixed readout duration (motivated by the desired spatial resolution, or a particular acquisition) wideband SSFP increases the null-to-null spacing in the steady-state frequency response, thereby reducing and potentially avoiding banding artifacts. This has been successfully applied to cine ventricular function imaging and cartilage imaging at 3 Tesla.
Instead, for a fixed band spacing requirement (motivated by off-resonance in a region-of-interest) this approach increases the available readout duration. This can enable higher spatial resolution in applications such as 3T SSFP coronary artery imaging, and can permit the use of time-efficient k-space trajectories such as echo-planar (9, 10) or spiral (11), which have been used for cardiac imaging at 1.5T.
Wideband SSFP may also be appropriate for clinical applications that fundamentally cannot be achieved using multiple-acquisition methods. MR fluoroscopy (continuous imaging) is one such example that requires a single steady-state that is free of banding artifacts. Another example is cardiovascular imaging, where flow and motion through bands cause transient artifacts.
There are also several possible extensions to this method. The signal and contrast during TRs is stronger than during TR, and therefore could be useful for improving overall SNR, CNR, and image quality. Low spatial frequency information (which requires shorter readout time) could be acquired during TRs and somehow combined with full spatial frequency information acquired during TR. It may also be possible to acquire low resolution field maps, coil sensitivity maps, or navigators during the short TR, that can be used to accelerate images or compensate for artifacts.
RF heating is an important constraint for this sequence. Wideband SSFP uses more RF pulses than conventional balanced SSFP for the same amount of readout time. To maintain SAR while using high a-values, the flip angle should be reduced, which may lead to suboptimal SNR or CNR. A possible solution is to use variable rate selective excitation (VERSE) (20) pulses to reduce SAR and permit the use of higher flip angles. If no data is collected during the short-TR, it may also be possible to design RF pulses that combine the −α and α (separated by TRs) into one composite pulse with lower total SAR than the two separate pulses.
Accelerating the approach to steady state will be important when combining this technique with any type of contrast preparation (e.g. fat saturation or inversion recovery). While a simple method is described in this manuscript, more accurate methods designed for conventional balanced SSFP (15, 21, 22) may be adapted for use with wideband SSFP. We speculate that magnitude transients during the approach to steady state will be shorter in wideband SSFP compared to conventional balanced SSFP because the magnitude of the steady-state magnetization is closer to Mo (see Fig. 2).
SSFP sequences, in general, are highly sensitive to unbalanced phase. For this reason, phase-encode pairing is used to minimize unbalanced phase caused by gradient distortions (23), and first-moment nulling is used when imaging in areas of flow (11). Although not fully explored at this time, wideband SSFP and alternating-TR SSFP in general may be more sensitive than conventional SSFP to unbalanced phase due to flow and gradient distortions.
The band-spacing and SNR efficiency of wideband SSFP are influenced by several parameters including the flip angle, a-value, TR, and tissue relaxation times. Optimal tradeoffs may be found experimentally, or by using Bloch simulation (see Fig. 6). When comparing wideband SSFP with conventional single-acquisition SSFP, wideband SSFP is expected to have longer scan-time, and a reduction in pass-band signal during TR, which in some cases will be offset by an increase in available readout duration and/or the use of the echo during TRs. When comparing wideband SSFP with multiple-acquisition SSFP using phase-cycling, wideband SSFP is expected to have shorter scan-time and lower SNR efficiency.
We have demonstrated a new technique, called wideband SSFP, that widens the band spacing in single-acquisition balanced SSFP imaging. Simulations indicate that wideband SSFP preserves high SNR efficiency and T2/T1 contrast. Phantom studies verified the expected band spacing increase in both available echoes. This technique has been successfully applied to the avoidance of banding artifacts in cine cardiac imaging and high-resolution knee imaging at 3T.
Wideband SSFP has a wide range of potential applications. Broadly stated, it enables the use of longer readout durations with a fixed band spacing (for improved spatial resolution and efficient k-space sampling), and enables an increase in the band spacing for a fixed TR (for the avoidance of banding artifacts). Wideband SSFP may be particularly useful for high-resolution balanced SSFP imaging at high-field strengths, due to the increased off-resonance and long required readout durations.
We thank Zungho Zun, Jon-Fredrik Nielsen, Garry Gold, and Gerald Pohost for discussions and collaboration. HLL has received the support of a USC Viterbi School of Engineering Graduate Fellowship.