Diffusion weighted image domain Propeller EPI (DW iProp EPI)

Introduction Geometric distortions in Echo Planar Imaging (EPI) are dependent on two sequence parameters: the phase encoding FOV and the echo spacing between two consecutive ky lines, both of which affect the k-space traversal speed. Assuming that a maximum readout gradient bandwidth is used, the echo spacing can be reduced by decreasing the resolution in the kx (not ky!) direction. Both Short Axis Propeller (SAP)-EPI and Readout-Segmented (RS)-EPI exploit this feature. The most common way to reduce the phase encoding FOV is with parallel imaging, where an R times lower phase FOV is acquired and unfolded to the desired full FOV using SENSE or GRAPPA. This leads to a corresponding distortion reduction by a factor of R. The phase encoding FOV can also be reduced for certain applications like spine imaging, where the final desired FOV is rectangular. For this case, a tilted or orthogonal refocusing pulse or a 2D spectral spatial RF pulse can be used to only excite the area of interest and avoid aliasing from tissue outside the FOV [1-4]. In this work we present a new pulse sequence for diffusion imaging, called image domain Propeller EPI (iProp-EPI). Here, we acquire a blade in the image domain with a reduced FOV, with successive blades acquired in subsequent TR's. The final isotropic FOV image is reconstructed by gridding the blades together in the image domain. This is different to other propeller-driven diffusion pulse sequences, such as PROPELLER and SAP-EPI, whereby blades are defined in k-space. iProp-EPI has some useful advantages for brain diffusion applications: (1) Similar to Zonally magnified (ZOOM)-EPI [4], geometric distortions are reduced by an amount corresponding to the FOV ratio between the frequency and phase encoding directions; (2) Since the overlap of blades occurs in the image domain, Nblades averages are obtained in the center of the brain – the most SNRstarved region for 'many'-channel coils (and often of special interest in deep WM DTI studies); (3) Even if each blade covers only a strip of the final image FOV, there is enough overlap between adjacent blades to make motion correction straightforward; (4) Since the image phase of the blade does not need to be preserved when combining the blades in the image domain, the gridded image is immune to potential spatially-varying non-linear phase changes, such as seen in DWI.