Introduction: Diffusion imaging is one of the most important applications for magnetic resonance imaging(MRI)since it is only feasible noninvasive technique for in vivo measurement of the microscopic motion of water in tissues.Multishot EPI(msEPI)is an alternative method,instead of clinical single shot EPI,which offers significant advantages in obtaining high-quality diffusion images: high spatial resolution,reduced susceptibility artifact,less eddy-current-induced distortion,and less blur.The major challenge of msEPI in diffusion MRI is that the image is typically corrupted by ghosting artifacts due to the phase variations from shot to shot.Methods: In this study,a K-space-based reconstruction method is proposed to produce high resolution diffusion images using navigated interleaved multishot EPI.Each subset of K-space data acquired using an msEPI was treated as a virtual multi-channel undersampling dataset,which can be reconstructed by generalized autocalibrating partially parallel acquisitions(GRAPPA)algorithm(a typical K-space method used in MRI for accelerated parallel imaging).The motion-induced phase errors were compensated by directly employing the navigator echoes as auto-calibrating signals(ACS)data in GRAPPA reconstruction.Isotropic diffusion-weighted imaging(DWI)of brain were acquired from healthy volunteers on an 1.5T MRI Scanner using a dual refocusing spin-echo(SE)diffusion-weighted msEPI sequence(as shown in Figure 1)with acquisition parameters as follows: TR/TE = 3200 ms/92 ms,matrix size = 256 x 256,FOV = 240 x 240 mm2,number of slices = 15,slice thickness = 6 mm.An interleave factor of 8 was used in msEPI,the echo train length(ETL)of both image-echo and navigator-echo was 32.Three orthogonal diffusion gradient direction imaging with b-value = 1000 s/mm2 was performed with three averages for each direction,and the final image was combined using the geometric mean of three diffusion direction images.Results: An 8-shot diffusion MRI example is shown in Figure 2.The first row shows the images directly reconstructed via fast Fourier transform(FFT)without motion correction,severe ghosting artifacts are remained due to the random motion-induced phase errors.The second row shows the results using the proposed method,that the phase errors are naturally compensated by using navigator-echo signals as ACS in GRAPPA reconstruction.No residual ghosting artifact is visible in these images.From left to right columns,images are diffusion encoded along RO,PE,and SS directions,and the rightmost column shows isotropic diffusion weighted images combined from the left three images of geometric means.To demonstrate the robust of the proposed method,more results of the isotropic DWI images are shown in Figure 3.Conclusions: A simple and convenient msEPI reconstruction method implemented in K-space for diffusion MRI is presented.Isotropic diffusion-weighted human brain imaging of healthy volunteer demonstrates that the proposed method provides robust reconstruction with good quality of high spatial resolution,no obvious residual artifact,and no apparent distortion.