Researchers and clinicians interested in high resolution diffusion imaging.Multi-shot acquisition strategies have been used to obtain images with high spatial resolution for diffusion weighted imaging (DWI), but they are susceptible to motion-induced phase variations among different shots ¹. In recent years, several techniques have been proposed to be capable of correcting shot-to-shot phase variations of multi-shot acquisition, including image domain methods ^2,3 and k-space domain methods ^4,5. However, these approaches result in blurred images and artifacts in presence of bulk motion of the subject. To address this problem, AMUSE ⁶ based on MUSE ², one of typical image domain methods, has been developed. In this work, we describe a new approach to reduce bulk motion induced errors as an extension of the SYnergistic iMage reconstruction using PHase variatiOns and seNsitivitY (SYMPHONY) ³, a k-space reconstruction method for 2D navigated multi-shot DWI, in order to obtain high spatial resolution diffusion images in presence of bulk motion.

Theory The large-scale motion assumedly is the in-plane rigid motion (translation and rotation) occurred between interleaved acquisitions. The proposed method is divided into three steps: motion parameter estimation, motion correction and reconstruction as illustrated in Fig. 1. In the first step, the rotation angle relative to a reference shot is estimated by finding the angle with maximum correlation between the reference and the target shot, using the circular k-space region of the navigator center ⁷. Similarly, parameters of translation are also estimated using the navigator. In the second step, the linear phase shifts in k-space due to translation are directly removed. Since the rotation in the image domain corresponds to the rotation in k-space, rotation of the k-space with estimated angles is required to correct the errors. Instead of direct rotation of the k-space data which will lose high frequency signals, we choose to rotate the k-space sampling trajectory. In the third step, non-Cartesian SPIRiT-based ⁸ SYMPHONY is implemented on the corrected k-space data and the sampling trajectory. It extends the SPIRiT interpolation from coil dimension to shot-coil dimension, based on the theory that the k-space of different shots are encoded by phase variations ³. Finally, artifact-free images of all shots are reconstructed and summed into a final diffusion image. The original SYMPHONY is implemented for comparison with the proposed method.

Simulation Motion simulations were designed to evaluate the proposed method. A 32-channel non-diffusion weighted 8-shot EPI image was acquired from a healthy volunteer on a Philips 3T scanner (Philips Healthcare, Best, The Netherlands). To simulate large-scale motion, image of each shot was randomly rotated (-15~15°) and translated (-10~10 pixels) in the x and y direction. Then, spatially random phases (third-order) were added to the 8-shot data respectively, to simulate motion-induced phase variations in diffusion weighted images. The matrix size of the data was 240×232, navigator size was 240×25. 32 channels were compressed to 6 channels while 99% of information was preserved ⁹.

In-vivo motion In-vivo head motion experiments were performed to validate the effectiveness of the proposed method to correct rigid motions. The volunteer rotated his head about ±8° every 15s~25s during the acquisition. The multi-shot diffusion weighted images were acquired with the following parameters: number of shot=8, FOV=240×240 mm², slice thickness=4 mm, TR/TE=3000/77 ms, in-plane image resolution=1×1 mm², the number of diffusion directions=3, b value=800 s/mm², navigator size is 89×29. Image coregistration between non-DW and DW images are used as corrections for both methods.

Fig. 2 shows the reconstructed images by direct Fourier transform, the original SYMPHONY and the proposed method in simulation studies. The original SYMPHONY reduced motion-induced phase variations but resulted in severe blurring due to large-scale motion. By contrast, both motion-induced phase variations and bulk motion-induced blurring were removed by the proposed method. Fig. 3a shows the reconstructed 8-shot diffusion images of two slices with one diffusion direction in the in-vivo motion experiment, and Fig. 3b shows the corresponding averaged diffusion weighted images (3 diffusion directions). The proposed method removed most artifacts compared with the original SYMPHONY.The simulations and in-vivo motion experiments validated the effectiveness of the proposed method to remove both the ghosting artifacts from minuscule motion and the blurring from macroscopic motion. It can provide high-resolution diffusion weighted images with a large number of shots (8-shot in brain). The further development of the proposed method to deal with more complicated motion such as through-plane motion will be the focus of future research efforts.