Development and validation of 3D MP-SSFP to enable MRI in inhomogeneous magnetic fields

Abstract

Purpose: We demonstrate the feasibility of MRI with missing-pulse steady-state free precession (MP-SSFP) in a 4T magnet with artificially degraded homogeneity.

Methods: T₁ , T₂ , and diffusion contrast of MP-SSFP was simulated with constant and alternate radiofrequency (RF) phase using an extended phase graph. To validate MP-SSFP performance in human brain imaging, MP-SSFP was tested with two types of artificially introduced inhomogeneous magnetic fields: (1) a pure linear gradient field, and (2) a pseudo-linear gradient field introduced by mounting a head-gradient set at 36 cm from the magnet isocenter. Image distortion induced by the nonlinear inhomogeneous field was corrected using B₀ mapping measured with MP-SSFP.

Results: The maximum flip angle in MP-SSFP was limited to ≤10° because of the large range of resonance frequencies in the inhomogeneous magnetic fields tested in this study. Under this flip-angle limitation, MP-SSFP with constant RF phase provided advantages of higher signal-to-noise ratio and insensitivity to B₁⁺ field inhomogeneity as compared with an alternate RF phase. In diffusion simulation, the steady-state magnetization in constant RF phase MP-SSFP increased with an increase of static field gradient up to 8 to 21 mT/m depending on simulation parameters. Experimental results at 4T validated these findings. In human brain imaging, MP-SSFP preserved sufficient signal intensities, but images showed severe image distortion from the pseudo-linear inhomogeneous field. However, following distortion correction, good-quality brain images were achieved.

Conclusion: MP-SSFP appears to be a feasible MRI technique for brain imaging in an inhomogeneous magnetic field.

Keywords: brain imaging; inhomogeneous magnetic field; missing pulse steady-state free precession.

Figure 1.

A) Sequence diagram of 3D MP-SSFP in inhomogeneous magnetic fields (ΔB₀). The sequence is composed of an RF pulse train with a constant interval, τ. Refocusing echo signals are acquired by replacing an RF pulse in every N_MP (=5) RF pulses with a signal acquisition; therefore, time from the refocusing echo peak to the center of the first RF pulse in the following TR is τ (= τ₁ + τ₂). Gradients can be applied during excitation (light blue) and/or readout (red) to compensate the inhomogeneous field. Phase encoding is performed along two dimensions for 3D imaging. B) Extended phase graph for MP-SSFP.

Figure 2.

Steady-state magnetization of MP-SSFP with constant (left column) and alternate RF phase (right column). A,B) Echo signals reaching the steady-state (9° flip angle for constant (A) and alternate (B) RF phase MP-SSFP). The refocusing echo signal intensities quickly increase at the beginning of the sequence and gradually reach the steady-state. As the missing pulse interval N_MP (TR=N_MP·τ) increases, stronger steady-state transverse magnetization is observed for both constant and alternate RF phase MP-SSFP. The steady-state magnetization is dependent on the flip angle (C,D). E,F) Increasing the steady-state magnetization by increasing N_MP is a more time-efficient way to improve SNR than increasing the number of averages. The solid line shows SNR improvement as a continuous function proportional to $\sqrt{T}$, as if averaging. The colored data points show normalized SNR compared to SNR with N_MP = 3 for N_MP = 3–9. The SNR advantage is more prominent with constant (E) RF phase than with alternate (F) RF phase. Simulations were conducted assuming a static gradient field of 7.6 mT/m, T₁/T₂ = 80/1200 ms and ADC = 1.0×10⁻³ mm²/s. Signal intensities were normalized with a spoiled gradient echo signal calculated with the Ernst equation (48) with TR=20 ms (5τ), flip angle = 9° and TE = 0 ms (T₂* decay was ignored).

Figure 3.

T₁ and T₂ contrast in MP-SSFP. A,B) Image contrasts are mostly T₁ dependent with constant RF phase MP-SSFP; with FA = 9°, short T₂ spins (<80 ms) showed stronger T₂ dependence. C,D) With alternate RF phase MP-SSFP, image contrasts are relatively poor with FA = 9°, whereas signal intensities with FA = 30° are roughly dependent on T₁/T₂. Simulations were performed with N_MP = 5, static gradient field = 7.6 mT/m and ADC = 1.0×10⁻³ mm²/s.

Figure 4.

Diffusion contrasts due to inhomogeneous fields for MP-SSFP with constant (A-C) and alternate RF phase (D-F). To see dependence on pulse interval τ, simulation was performed with τ = 4 (A,B) and 3 ms (D,E). Interestingly, the highest steady-state signal intensity increases with constant RF phase MP-SSFP as the field gradient increases up to around 16 and 21 mT/m for τ = 4 and 3 ms, respectively (A,B). Signal intensity ratio of τ = 3 and 4 ms (S_{τ=3 ms} /S_{τ=4 ms}) demonstrates that reduced diffusion weighting with τ = 3 ms provides an SNR advantage with low flip angles as compared to τ = 4 ms (S_{τ=3 ms} /S_{τ=4 ms} > $\sqrt{3/4}$), particularly with strong background fields (C). With alternate RF phase MP-SSFP, the steady-state magnetization is highest with no background field independent of τ and decreases gradually as the field gradient increased (D,E). The SNR advantage with τ = 3 ms is more conspicuous in MP-SSFP with alternate RF phase than with constant RF phase (F). Simulations were performed with N_MP = 5, T₁/T₂ = 1200/80 ms and ADC = 1.0×10⁻³ mm²/s.

Figure 5.

Experimental validation of diffusion weighting. Simulations were conducted with matched T₁, T₂ and ADC values with an agar gel phantom (T₁/T₂ = 2500/80 ms and ADC = 2.0×10⁻³ mm²/s) for MP-SSFP with constant (A) and alternate (B) RF phase settings. Comparison of steady-state magnetization from simulation and experiments at 4 T for static field gradients of 3.7, 7.6 and 15.3 mT/m (C,D). Similar to the simulation result in Fig.4A, simulation with constant RF phase MP-SSFP showed an increase of the steady-state magnetization up to a static field gradient of 7.8 mT/m. Consistent results were observed in experiments. Similarly, the simulation results matched the experimental measurements well with alternate RF phase MP-SSFP. For lower flip angles up to 7°, the constant RF phase setting provided higher steady-state signal intensities as compared to the alternate phase setting.

Figure 6.

B₁⁺ field dependence of the MP-SSFP steady-state magnetization. A) A B₁⁺ field of a TEM head coil was measured with the actual flip angle imaging (AFI) method at 4 T (left: sagittal and right: axial). FDA guidelines limited flip angles to under 10° due to the inhomogeneous magnetic fields used in this study. At low flip angle (nominally 6°), B₁⁺ field inhomogeneity induces a larger variation in the steady-state magnetization using alternate RF phase MP-SSFP than it does with constant RF phase MP-SSFP (B,C). Regions with weak B₁⁺ fields showed stronger steady-state signals with constant rather than alternate RF phase MP-SSFP (D).

Figure 7.

A-F) In vivo human brain imaging with MP-SSFP in a static gradient field of 15.1 mT/m along the anterior-posterior (y) axis. Sequence parameters were: τ=3.2 ms, N_MP = 5 (TR=16 ms), FA = 6.6°, HS2 pulse with a 91 kHz bandwidth and TA=3 min 43 s.

Figure 8.

Image distortion correction using a B₀ map obtained with MP-SSFP for a pseudo-linear field generated by mounting a head gradient 36 cm away from the 4 T magnet center. Spatial encoding was performed with phase encoding gradients along three spatial dimensions such that the reconstructed image is free from distortion from the pseudo-linear magnet field (A). A measured B₀ map (B) was fitted with 4th degree polynomials (C). 3D MP-SSFP using the pseudo-linear field for frequency encoding showed conspicuous intensity and geometric distortion (D). Such image distortion was corrected using the measured B₀ map (C); the distorted edge of a Teflon object in a phantom (arrows) is improved to yield a straight line in the corrected image (E).

Figure 9.

In vivo human brain imaging with MP-SSFP in the pseudo-linear magnetic field. The pseudo-linear magnetic field induced severe image distortion (A). The image distortion was corrected using the B₀ map shown in Fig.9C (B). The image shows anatomical structures with T₁, T₂ and diffusion contrasts (C-E). A region around the top of the brain shows a signal drop due to strong diffusion weighting associated with the steep field gradient in this area. Sequence parameters were: τ = 2.6 ms, N_MP = 6 (TR = 15.6 ms), FA = 4°, HS2 excitation with a 140 kHz bandwidth and TA = 5 min 30 s.