Reduced field-of-view (rFOV) imaging can provide vast benefits and flexibility in terms of acquisition time and spatio-temporal coverage, when the region-of-interest (ROI) is smaller than the full FOV (fFOV).^1–6 For gradient-echo imaging, the only viable rFOV method at short repetition times (e.g. 3D imaging, EPI) is through selective multi-dimensional (2D/3D) RF excitation (SELEX).^7–12,14 Although small tip angle solutions exist,^3,7,8 rFOV imaging benefits from large tip angle (LTA) excitation, and many acquisitions even require accurate LTA excitation.¹³ The methods proposed till date^9-12,14 are either approximations,¹¹ or non-unique,¹⁰ or computationally involved.^10,12 Most methods typically yield excitation pulses with low time-bandwidth product (TB < 10), leaving them vulnerable to B₀ inhomogeneity-related signal loss and distortion at high fields. Here, we propose to use the versatile field shaping capability of the dynamic multi-coil technique (DYNAMITE),^15–18 in conjunction with 1D-Shinnar-Le Roux (SLR) pulses¹⁹ to obtain 2D SELEX. We show that this combination can yield highly selective (TB > 12) 2D excitation.

We identified mouse brain imaging (fast 3D T₁ and T₂ mapping¹³ and EPI) as potential applications for this technique. Previously described DYNAMITE hardware^15,16 comprising of 48 coils was used for the studies. The main challenge in the proposed method was to meaningfully link the DYNAMITE field distribution to the 1D SLR pulse to be designed for excitation. Since SLR pulse design inherently links bandwidth (BW), transition width (TW), pass-and stop-band ripples and the pulse TB,¹⁹ we used an adaptive pulse design approach. The coil currents that produce the desired B₀ field and the corresponding excitation pulse were concurrently optimized to maximize SELEX of the brain.

A transition zone was defined¹⁶ around the mouse brain ROI (Fig.1A). The ROI spanned 9x7.5x10 mm in x (phase encode), y (phase encode 2) and z (readout) directions. In the transition zone, varying B₀ gradients (100 to 400 Hz/pixel) were prescribed separately (Fig.1A) to identify a starting point for the optimization of SELEX coil currents. For a given field distribution, constrained-least-squares fitting was performed at $$$\pm$$$1 A dynamic range to estimate currents that minimized the errors in the fit to the field distribution. Simultaneously, the fit restricted the fields generated within the ROI to positive values and forced the fields in the region of non-interest (RON) to negative values (Fig.1A). From this partitioned field distribution, the BW in the ROI and TW in the transition zone were used to design a suitable minimum (or linear) phase 90^o SLR excitation pulse. The pass- and stop-band ripple were limited to 1%. Bloch-equation simulations provided off-resonance behaviour of the pulse, which was used to assess the objective function $$$\sum_{ROI}sin(flip angle)/\sum_{RON}sin(flip angle)$$$. After identifying suitable starting coil currents for SELEX, the restrictions on the DYNAMITE fields generated in the transition zone was removed and the objective was maximized using a local optimizer. The optimization also yielded the corresponding SLR pulse and the required transmitter frequency offset for SELEX. The process was fully automated, requiring only ROI prescription. Proof-of-principle fast 3D gradient-echo experiments were performed in two euthanized mice at 9.4 T. The DYNAMITE hardware was also used for background B₀ shimming, including during excitation.

Successful 2D SELEX could be achieved in the experiments, with TB > 12 SLR pulses.The predicted and measured excitation patterns matched closely (Fig.1C,G). The FOV reduced from 15x15 mm to 9x12 mm in the two phase encoding dimensions. For isotropic acquisitions, this translates to acquisition acceleration of 2.1. The measured rFOV/fFOV signal ratio (Fig.1H) had Mean $$$\pm$$$ SD = 0.987 $$$\pm$$$ 0.069. The ratio in the RON had Mean $$$\pm$$$ SD = 0.0386 $$$\pm$$$ 0.023, indicating near absence of excitation. Images obtained from an accelerated rFOV excitation experiment are shown in Fig.2, and show no signs of aliasing. The coronal and sagittal fFOV and SELEX results are shown in Fig.3.The DYNAMITE approach simplifies the difficult problem of multi-dimensional SLR excitation to the familiar 1D SLR excitation case through its versatile B₀ field shaping capability. In combination with background B₀ shimming, it is a powerful tool for rFOV excitation, especially due to the high TB achievable. A unique attribute of rodent brain imaging which leads to high TB pulses is that, using DYNAMITE, SELEX of open-sided 2D problems are more amenable than fully enclosed regions. Inclusion of the skin above the skull in the ROI significantly increased the achievable TB, compared to restricting the ROI to the brain. Increasing coil current dynamic range from 1 A to e.g. 5 A is expected to mitigate this, and provide high TB SELEX over fully closed ROIs.