Gradient-Echo & Steady-State Imaging

Why Reverse a Gradient Instead of Flipping the Spins?

In a spin-echo sequence, a 180 degree refocusing pulse reverses the dephasing caused by static field inhomogeneity, so the measured decay follows the true $T_2$. Gradient-echo (GRE) sequences deliberately omit that refocusing pulse. The echo is formed instead by reversing the polarity of a readout gradient: a negative lobe first dephases the spins along the frequency-encode axis, then a positive lobe rephases them, producing a measurable signal peak at the moment the net gradient area returns to zero.

This single design choice cascades into everything that makes GRE the workhorse of fast, dynamic and three-dimensional imaging. Because no 180 degree pulse is required, the echo time $TE$ can be very short and the repetition time $TR$ can be pushed down to a few milliseconds. Because static inhomogeneity is not refocused, contrast reflects $T_2^*$ rather than $T_2$, making GRE exquisitely sensitive to susceptibility effects. And because each excitation typically uses a small flip angle, longitudinal magnetization is largely preserved, enabling rapid repeated sampling without long recovery delays.

The GRE Signal Equation

Consider a sequence that applies an RF pulse of flip angle $\alpha$ once every $TR$ and samples an echo at time $TE$. After many repetitions the longitudinal magnetization reaches a steady state: the regrowth between pulses exactly balances the fraction tipped away on each excitation. For a spoiled sequence (where any leftover transverse magnetization is destroyed before the next pulse) the steady-state signal is given by the classic FLASH/SPGR expression.

Read the three factors in order. The $\sin\alpha$ term is the transverse component created by the pulse. The fraction in the middle is the steady-state longitudinal magnetization available before each pulse, balancing recovery (numerator) against incomplete recovery from the previous tip (denominator). The final exponential is the $T_2^*$ decay accrued during $TE$. Together they explain why, at short $TR$, simply increasing the flip angle does not monotonically increase signal: a large $\alpha$ creates more transverse signal per pulse but leaves less longitudinal magnetization to recover, and at short $TR$ that magnetization cannot fully regrow.

The Ernst Angle

Since signal rises and then falls with flip angle, there is an optimum. Ignoring the $T_2^*$ term (which is independent of $\alpha$), we maximize $S(\alpha)$ by setting $\partial S / \partial \alpha = 0$. Differentiating the product $\sin\alpha \, / (1 - \cos\alpha \, E_1)$ with $E_1 \equiv e^{-TR/T_1}$ and setting the numerator of the derivative to zero gives the condition $\cos\alpha_E = E_1$.

The Ernst angle is the flip angle that delivers maximum steady-state signal for a given $TR/T_1$ ratio. Two limits are instructive. When $TR \gg T_1$, the exponential approaches zero, $\arccos(0) = 90$ degrees, and full recovery favors a hard pulse, exactly as in conventional long-$TR$ imaging. When $TR \ll T_1$, the exponential approaches 1 and $\alpha_E$ becomes small. A useful small-angle approximation is $\alpha_E \approx \sqrt{2 \, TR / T_1}$ in radians.

| TR/T1 | exp(-TR/T1) | Ernst angle alpha_E | | --- | --- | --- | | 0.005 | 0.995 | approx 5.7 degrees | | 0.02 | 0.980 | approx 11.5 degrees | | 0.05 | 0.951 | approx 18 degrees | | 0.10 | 0.905 | approx 25 degrees | | 0.50 | 0.607 | approx 53 degrees | | 2.0 | 0.135 | approx 82 degrees |

Spoiling: From Residual Coherence to Clean T1 Weighting

Equation 6.1 assumes that no transverse magnetization survives from one $TR$ to the next. In reality, with short $TR$ the transverse component has not fully decayed when the next pulse arrives, and that residual coherence can refocus into unwanted stimulated echoes that corrupt $T_1$ contrast. Spoiling removes it. Two mechanisms are combined in practice.

1. Gradient spoiling: a strong gradient lobe at the end of each $TR$ imposes a large spatially varying phase, so that summed over a voxel the residual transverse magnetization cancels. Used alone it is imperfect because the dephasing pattern repeats each $TR$.
2. RF spoiling: the phase of the excitation pulse is incremented by a quadratically growing amount each $TR$ (a common increment is 117 degrees). This pseudo-randomizes the phase of any carried-over coherence so it averages to zero, yielding signal that closely follows the ideal spoiled equation.

Properly spoiled GRE goes by vendor names SPGR (GE), FLASH (Siemens) and T1-FFE (Philips). With a short $TE$ to suppress $T_2^*$ and a flip angle at or above the Ernst angle to emphasize $T_1$ differences, spoiled GRE produces $T_1$-weighted images very quickly. Its three-dimensional form is the foundation of MPRAGE (magnetization-prepared rapid gradient echo), where an inversion pulse precedes a fast GRE readout train to give high-resolution $T_1$-weighted volumetric brain imaging.

Balanced SSFP: Keeping the Magnetization

Spoiling throws away transverse coherence. The opposite strategy is to preserve and reuse it. In balanced steady-state free precession (bSSFP) every imaging gradient on all three axes is fully rewound within each $TR$, so the net gradient area per $TR$ is zero on every axis. Combined with a phase-alternated RF train (typically $\alpha, -\alpha, \alpha, \dots$), the transverse magnetization is refocused and carried forward, building a remarkably high steady-state signal. Vendor names are TrueFISP (Siemens), FIESTA (GE) and balanced-FFE (Philips).

Because both longitudinal and transverse magnetization are recycled, the on-resonance bSSFP signal in the short-$TR$ limit depends on the ratio $T_2/T_1$ rather than on $T_2^*$ decay.

This $T_2/T_1$ dependence is why fluids (blood, CSF, bile) with high $T_2/T_1$ appear bright on bSSFP, giving the characteristic look of cardiac cine, fetal and MR-cholangiopancreatography-style imaging. The SNR efficiency of bSSFP is the highest of any GRE family because virtually no signal is discarded. The cost is banding artifact: when off-resonance phase accrual per $TR$ approaches 180 degrees, the refocusing condition fails and dark bands appear. Keeping $TR$ short (for example 3 to 4 ms) widens the spacing of these bands, and good shimming pushes them out of the field of view.

| Property | Spoiled GRE (SPGR/FLASH) | Balanced SSFP (TrueFISP/FIESTA) | | --- | --- | --- | | Residual transverse Mxy | Destroyed (spoiled) | Refocused and reused | | Dominant contrast | T1 (with T2* via TE) | Mixed, follows T2/T1 | | Relative SNR efficiency | Moderate | Highest in GRE family | | Main artifact | Susceptibility / T2* dropout | Off-resonance banding | | Typical use | Dynamic T1, 3D, MPRAGE | Cardiac cine, fluid-bright, fetal |

Explore the Sequence Timing

The diagram below lets you step through the RF, slice, phase and frequency-encode lines of a gradient-echo sequence and watch how reversing the readout gradient creates the echo. Note the bipolar readout lobe (negative then positive) and, for the spoiled variant, the spoiler gradient at the end of $TR$.

Track the area under the readout gradient: the echo peaks exactly when the cumulative gradient area since the dephasing lobe returns to zero. Shortening $TR$ compresses the whole diagram, which is precisely what enables sub-second 2D frames and breath-hold 3D volumes.

Chemical Shift of the Second Kind: In-Phase and Opposed-Phase

Because GRE does not refocus the chemical-shift frequency difference between water and fat, the relative phase of the fat and water signals evolves freely during $TE$. Fat precesses slower than water by approximately $3.5$ ppm, which corresponds to a frequency difference $\Delta f = 3.5 \times 10^{-6} \, \gamma B_0 / 2\pi$ of about $220$ Hz at $1.5$ T and about $440$ Hz at $3$ T. The two signals therefore drift in and out of phase as $TE$ increases. This is chemical shift of the second kind, distinct from the spatial misregistration (first kind) seen along the frequency axis.

At $1.5$ T the signals are opposed-phase near $TE \approx 2.3$ ms and in-phase near $4.6$ ms; at $3$ T these halve to roughly $1.15$ ms and $2.3$ ms. In a voxel containing both water and fat, the opposed-phase image shows signal cancellation, and the boundary between a fat-containing organ and adjacent water-rich tissue acquires a black outline, the so-called India-ink artifact, which confirms a true fat-water interface.

Play With GRE Contrast

Use the contrast playground to vary $TR$, $TE$ and flip angle and watch how the GRE signal and tissue contrast respond. Push the flip angle through the Ernst angle for a fixed short $TR$ and observe the peak-then-fall behavior; then lengthen $TE$ to bring out $T_2^*$ dephasing and susceptibility effects.

Why GRE Underpins Modern Protocols

Short $TR$, short $TE$ and small flip angles make GRE the natural engine for anything that must be fast or volumetric: breath-hold abdominal $T_1$ imaging, dynamic contrast studies, time-resolved MR angiography, cardiac cine, real-time and 4D-flow acquisitions, and the 3D $T_1$ anatomic backbones (MPRAGE, VIBE/LAVA/THRIVE) used across the body. The two extremes of coherence handling, full spoiling for clean $T_1$ weighting and full balancing for high-SNR $T_2/T_1$ contrast, plus the inherent $T_2^*$ sensitivity, give the radiologist a toolkit that spans from susceptibility mapping to bright-blood fluid imaging, all built on the simple act of reversing a gradient instead of flipping the spins.