Single-shot SPEN MRI is a technique capable of retaining the time efficiency of single-shot EPI but with significantly reduced geometric distortions. Akin to EPI, the phase inconsistency between even and odd echoes also results in Nyquist ghosts in SPEN images. However, the characteristic of the SPEN signals provides the convenience of obtaining two alias-free images directly from solely even and odd echoes without additional reference scans. The phase difference of the two images can thus be utilized to implement two-dimensional Nyquist ghost correction. This work presents a scheme with more reliable performance than the one previously reported ¹. The phase difference map is utilized to correct the encoding matrix instead of the SPEN signals, avoiding the modulation possibly introduced on the profile during so-called back and forth transformation, and removing ghosts without blurring the images.

The flowchart of the proposed scheme is listed in Fig.1 with more explanations as follows.

1. pFT stands for partial Fourier reconstruction ².

2. |I_evenI_odd*| was used as the weight for 2D polynomial fitting. When dealing with multichannel data, the fitting was implemented jointly on all channels to obtain a single fitted ∆Φ_fit ³.

3. The phase-corrected partial Fourier reconstruction was based on

$$I(x,y)=\sum_{n=n_{0}-N_{p}/2}^{n_{0}+N_{p}/2-1}{s(x,n)e^{-i(ay^2+by-2\pi n\triangle k_{y}y)}e^{-i\triangle \theta _{n}(x,y)}}$$

$$\triangle\theta_{n}(x,y)=\begin{cases}\triangle \Phi_{fit}& \text{if $n$ is even} \\0 & \text{if $n$ is odd} \end{cases}$$

where x and y are the coordinates of a given voxel, n₀ is the index of echo corresponding to the refocusing of the voxel, N_p is the number of echoes involved in reconstructing the voxel, s(x,n) are SPEN signals obtained after step 1, a and b are the quadratic and linear phase modulation coefficients imparted by the frequency-swept excitation, and Δk_y is the k-space step along SPEN dimension.

4. For multichannel imaging, the ghost corrected images of all channels were finally combined using root of sum of squares.

The effectiveness of the proposed ghost correction scheme was validated by experiments at 3T on Siemens Prisma scanner and at 7T on Varian animal scanner. Its robustness was further verified by distortion unwarping on the ghost corrected data at 7T.

Human brain experiments were conducted on four healthy volunteers with informed consent. Three orthogonal orientations were scanned with fat suppression, 24 slices for each. FOV=220×220mm² with slice thickness=5mm, single-shot SPEN MRI were scanned with TR/TE=5000ms/66 ms, acquisition matrix=64×64, SW=1502Hz/voxel, bandwidth/duration/flip angle of excitation=50kHz/5.12ms/30°, referential TSE images were scanned with TR/TE=5000ms/104ms and acquisition matrix=256×256. Figure 2 shows that single-shot SPEN MRI can provide images with high fidelity similar to the multi-shot TSE images, even at regions susceptible to distortions, such as those near orbital frontal/temporal/occipital lobes and spinal cord. The Nyquist ghosts are manifested as ghosts superimposed on the parent images and stripe-like artifacts which were successfully removed by the proposed phase-corrected pFT reconstruction without blurring the images, as is the case by SR method.

For in vivo rat experiments, the rats were anesthetized with chloral hydrate solution (10%) through intraperitoneal injection (0.3ml/100g) before scans were carried out. FOV=100×80mm² with slice thickness=2mm, single-shot data were acquired by self-refocused SPEN sequence with TR/TE=5000 ms/34 ms, acquisition matrix=64×64, SW=250kHz, bandwidth/duration/flip angle of excitation=9.8kHz/26.1ms/90°, referential multi-shot GRE image was scanned with TR/TE=20ms/3ms and acquisition matrix=128×128. The field map needed for unwarping was calculated from the phase difference of two scans, one was symmetric and the other was asymmetric with a time delay of 1ms inserted before the echo train. Phase-corrected pFT reconstruction was implemented to obtain the distorted source image with possible Nyquist ghost removed before subsequent distortion correction ⁴. Figure 3 shows that the Nyquist ghost correction procedure did not compromise the unwarping performance. For the torso slice, the position of the readout window was hard to be perfectly aligned with the echo train, giving rise to some stripes at the bottom of the image before unwarping which led to undesired artifacts and incomplete signal recovery around the locations after unwarping (marked). With ghost correction, the stripe artifact on the warped image was removed, improving the accuracy of the field map for subsequent distortion correction and thus providing a better unwarping result.

Experimental results demonstrate that the proposed technique can provide more reliable ghost correction than the previously reported measure without blurring the image. It is applicable for both single-channel and multi-channel acquisition. It would not compromise the performance of subsequent distortion correction at high-field, but help to obtain a better unwarping result when Nyquist ghost is present.