In standard echo-planar Imaging (EPI)¹, an oscillating trapezoidal readout (RO) gradient is combined with a blipped phase-encoding (PE) gradient, resulting in a rectilinear k-space trajectory, which can be easily combined with parallel imaging². No data are acquired during the blipped PE gradient because the trajectory would deviate from a Cartesian path at the edges of k-space in the RO direction. A significant reduction in sound pressure can be achieved by using a sinusoidal waveform³, which preserves the rectilinear k-space trajectory (Fig. 1A). Acoustic noise is reduced further by using a constant PE gradient³. However, this presents challenges for parallel imaging because the k-space trajectory is no longer Cartesian, but an imperfect zigzag (Fig. 1B). Data acquisition however can now take place continuously, allowing a more efficient use of the gradient system. In this paper we introduce a new EPI sampling scheme – the Zigzag-Aligned-Projections (ZAP) EPI - which adapts the constant PE scheme to give a true zigzag trajectory in k-space. Therefore the two advantages (higher efficiency and lower acoustic noise) of the constant PE EPI are maintained and can now be combined with standard parallel imaging techniques, such as GRAPPA². The new sampling scheme replaces the constant PE gradient with a low-amplitude variable gradient that is the modulus of the RO waveform to give a true zigzag trajectory (Fig. 1C). The data are processed with a 1D phase correction, a 1D interpolation for the variable density sampling along the zigzag and are finally time reversed . Due to time reversal, the data are now aligned on a Cartesian grid. However, phase errors of higher order appear between odd and even echoes . This will result in folding artifacts in PE direction. The zigzag sampled data can be reconstructed using an interlaced Fourier Transformation (iFT)⁴. In addition, for parallel imaging, the data can be separated into odd and even echoes to give two data sets with consistent phase, which can be processed using a fixed GRAPPA kernel as in the Cartesian-sampled case². Afterwards the data are combined to yield the final image. An image reconstructed using iFT is shown in Fig. 2A and an image reconstructed using GRAPPA with an acceleration factor of 2 is shown in Fig. 2B. Images were acquired from healthy volunteers using a Siemens MAGNETOM Prisma 3T system and acoustic noise measurements were performed on a Siemens MAGNETOM Skyra 3T system.

The image reconstructed with iFT is prone to artifacts . The GRAPPA reconstructed image shows no artifacts and this is therefore the preferred reconstruction technique. The acoustic noise measurements showed that the effect of the PE gradient of the ZAP sampling scheme on the overall acoustic noise is negligible and at the same level as the acoustic noise generated by the constant PE EPI scheme.

We have shown that a true zigzag k-space trajectory can be achieved with EPI by replacing the constant-amplitude PE gradient with a waveform which is the modulus of a sinusoid. The phase errors relating to the alternating RO gradient can be corrected by either an iFT or a GRAPPA-based reconstruction. Acoustic noise is on the same level as the constant PE EPI – regarded as the quietest EPI. Further measurements are planned to verify the effect of higher efficiency of the ZAP EPI readout compared to the standard blipped-PE approach to EPI.