In-plane reduced field-of-view (rFOV) excitation based on 2DRF is useful in many applications, such as to circumvent chest motion artifacts in cardiovascular MRI ¹, to achieve localized B₀ shimming with reduced SAR ², to reduce the signal loss induced by intravoxel dephasing in T₂^*W fMRI ³, and to target the anatomically pre-defined ROI in single-voxel MR spectroscopy for minimized partial-volume effects ⁴. The EPI-style gradient waveform is commonly used in 2DRF implementation. However, at high field, the off-resonance effects during excitation cause severe distortions of the profiles obtained. Single-shot spatiotemporally encoded (SPEN) MRI is capable of maintaining the time efficiency of EPI but with significantly improved robustness to off-resonance induced distortions ⁵. Similar mechanism can be applied to 2DRF implementation to obtain excitation profiles of higher accuracy ⁶. In this work, we investigate the feasibility to achieve in-plane selective excitation of anatomically pre-defined regions using SPEN-2DRF pulse under different shim conditions.In vivo rat brain imaging was performed on a 7T Varian animal scanner using two sequences depicted in Fig.1, in which the 2DRF pulses replace their 1D counterparts for exclusively exciting the brain regions. The 180° sinc pulses serve as slice selection and refocusing. The excitation and acquisition stages of both sequences are mirror symmetrical with identical gradient waveforms, magnitude and durations. The flowchart of the overall SPEN in-plane excitation scan is shown in Fig.2. Multi-shot GRE images were acquired for positioning and delineating the areas to be excited in subsequent rFOV single-shot scans. Compared with conventional method, the excited pattern of SPEN-2DRF is a profile with quadratic phase modulation. Partial Fourier reconstruction ⁵ was used to obtain the SPEN images. Scans were separately conducted after shimming and without shimming. The Fourier-based 2DRF and SPEN-2DRF pulses were both composed of 64 subpulses, and the latter had a time-bandwidth product of 256 along the SPEN dimension. Other parameters shared by the two sequences included FOV=60×60mm² (axial and coronal), and 80×80mm² (sagittal) with slice thickness=2mm, TR/TE=5000ms/34ms, durations for excitation and acquisition=26.1ms, acquisition matrix=64×64, and acquisition bandwidth=250kHz.Results of three orthogonal orientations were illustrated in Fig.3. For full FOV imaging, the images obtained with single-shot EPI were severely impaired by geometric distortions and local signal pileups due to shim inefficiency and susceptibility variations near tissue-air interfaces. In contrast, single-shot SPEN MRI provided images with significantly reduced deformation even without shimming. Correspondingly for rFOV imaging, the profiles obtained from spatiotemporally encoded 2DRF excitation and acquisition are of better geometric fidelity than those obtained from conventional method. Particularly, the shim conditions had a strong impact on the accuracy of the profiles obtained with conventional method, while the profiles obtained with SPEN method are largely immune to shimming conditions.In order to reduce the distortions caused by off-resonance effects during excitation and acquisition, the time-bandwidth product of the SPEN-2DRF pulse along the blipped (SPEN) axis should be several times larger than the number of subpulses, resulting in the periodic excitation of overlapped side-lobes. With small-tip-angle assumption, these excitation replicas share the same intensity and quadratic modulation, but differ by a linear phase. The undersampled acquisition along SPEN dimension also results in periodic aliases separated by the same length ∆L as the excitation replicas. With partial Fourier reconstruction, the value of a voxel is the sum of a series of overlapped side-lobes. Figure 4 shows the distribution of weights of all side-lobes, which can be divided into three levels. The second level is 10 times lower than the highest and the third is even smaller. Therefore, except for the four main lobes(5 for the center voxels only), the influence of other side-lobes can be neglected. Because the size of the rat head is small, only regions at the edge of the brain is overlapped by the nearest side-lobe in the results presented in Fig.3. For a larger object, such as the human brain, the signals of different lobes should be decomposed by further processing. Unlike conventional 2DRF excitation, these overlapped signals originate from different voxels separated by ∆L, providing a possibility to decompose them with parallel acquisition and reconstruction, and also the potential for applying this technology to clinical scanners not equipped with multi-channel transmission coils. Additionally, because the SPEN-2DRF pulse usually has a much larger duty cycle than Fourier-based 2DRF, the peak power needed is smaller.A novel spatiotemporally encoded in-plane selective excitation and acquisition method is proposed in this work. Experimental results of in vivo rat shows that it can produce profiles less susceptible to off-resonance induced distortions at high field.