Multi-shot EPI (ms-EPI) can achieve higher resolution DWI than single-shot EPI (ss-EPI), but the ghost artifacts caused by shot-to-shot phase variations must be corrected. Some k-space correction methods for ms-EPI DWI have been proposed^1,2,3. These methods explore the relationships among different shots, and use them for k-space interpolation in an extended GRAPPA⁴ (eGRAPPA) fashion to recover the unacquired data. However, these methods require sequence with additional navigator acquisition after the image echo, which prolongs the scan time. In this work, a self-calibrated method for multi-shot DWI correction in k-space is proposed, which does not require navigator acquisitions for efficient scanning and does not suffer from the mismatch between image and navigator echoes^1,2,5. This method is demonstrated with interleaved ms-EPI DWI in liver.

Reconstruction Recent works^1,2,3 report the motion induced phase variations can be treated as a power of encoding and k-space relationships among different shots are used for k-space eGRAPPA interpolation. In these studies, additional navigator echoes are needed to calculate the interpolation parameters, which prolongs the scan time by about 1/3. In this work, self-calibrated eGRAPPA is implemented as follows: 1) virtual navigators are firstly generated by recovering the missing data in each shot using the conventional GRAPPA⁴; 2) the calibration among shots is performed from these calculated navigators followed by eGRAPPA interpolation as in previous work¹; 3) partial Fourier reconstruction and complex image combination⁶ are performed to generate the final diffusion image. Fig. 1 shows the self-calibrated eGRAPPA reconstruction pipeline using 3-shot as an example (a), a simple demonstration of GRAPPA and eGRAPPA (b) and a simple comparison of navigated and navigator-free sequence (c).

Experiments Human liver DWI scans were performed on a Philips 3.0T MRI scanner (Philips Healthcare, Best, The Netherlands) using a 16-channel abdominal coil. Data were acquired from 2 healthy volunteers under IRB approval from our institution. A navigated ms-EPI sequence was scanned for comparing different reconstruction results with and without using the acquired navigators. The data were acquired using free-breathing with respiratory trigger, acquisition voxel size = 2 × 2 × 6 mm³, 4 shots with 22 echoes per shot, partial Fourier factor = 0.7, TE = 41 ms, scan time = ~3 min, 3 orthogonal diffusion directions with b = 500 s/mm², NSA = 4. For comparison, ss-EPI DWI (acquisition voxel size = 3 × 3 × 6 mm³) with GRAPPA = 2 was also scanned for references. Proposed self-calibrated eGRAPPA method was compared with navigator-calibrated eGRAPPA, image-domain self-calibrated MUSE⁷ and conventional GRAPPA-average reconstruction⁸. For the k-space interpolation kernel size, 3×3 was used in the first GRAPPA step and 3×5 in the second eGRAPPA step.

Fig. 2 shows the isotropic liver DWI using ss-EPI as a reference (a), ms-EPI without correction (b), with conventional GRAPPA correction (c), with proposed self-calibrated eGRAPPA correction (d), with navigator-calibrated eGRAPPA correction (e) and with MUSE correction (f). The phase errors cause severe ghost artifacts in ms-EPI if no correction is performed (b). The proposed correction method (d) can correct these ghost artifacts, and reduced noise level can be visually seen compared with conventional GRAPPA-average correction (c). Proposed self-calibrated eGRAPPA (d) and navigator-calibrated eGRAPPA (e) provide similar image quality in the shown slice. Image-domain MUSE result (f) also shows effective correction except for some residual artifacts (yellow arrow heads). This may be caused by the insensitivity to motion of eGRAPPA inherited from GRAPPA⁸, compared with SENSE-based method.

We further compared self-calibrated and navigator-calibrated eGRAPPA and chose three slices shown in Fig. 3. In these cases, although navigator-calibrated method can correct most of the ghost artifacts, proposed self-calibrated method shows better correction with fewer residual artifacts (yellow arrow heads). The reason can be that the self-calibrated strategy is not sensitive to the potential mismatch between image and navigator echoes due to different distortion levels, resolutions^1,2,5 (e.g. Fig. 1c), motion states or signal attenuation (T2 decay, non-CPMG).

The navigator-free ms-EPI sequence uses single-echo spin-echo acquisition, instead of navigated two-echo spin-echo sequence, in order to improve the scan efficiency. In the meantime, the image quality is not affected much (Fig.2 d,e), and sometimes even shows better correction (Fig. 3). However, like other two-step self-calibrated methods (MUSE, etc.), the number of shot to cover the full k-space is still limited in the first calibration step (e.g. no more than 6-shot using 8-channel coils), which is not the limitation for methods with navigator acquisitions.

The proposed self-calibrated eGRAPPA method can correct motion induced ghost artifacts in multi-shot diffusion imaging, which is efficient without the need for navigator acquisitions in sequences.