Irretrievable signal loss in partial-Fourier acquired diffusion-weighted images

Introduction Diffusion weighted (DW) imaging is highly sensitive for bulk motion, which frequently occurs in vivo, such as in the lower brain areas, where the parenchyma pulsates rhythmically due to cardiac activity. It is known that for partial-Fourier (PF) imaging, the emergence of pulsation artefacts in DW images depends critically on the image reconstruction method. Specifically, it has been shown that zero-padding PF reconstruction reduces pulsation artefacts compared to the Margosian method and that homodyne PF reconstruction reduces these artefacts compared to full-Fourier (FF) reconstruction. The current explanation is that cardiac pulsations give rise to high-resolution phase variations that are not accounted for by these PF methods’ low-resolution phase estimates. More advanced reconstruction methods have been developed that provide better (high resolution) phase estimates. However, all PF reconstruction methods suffer from pulsation artefacts when low-frequency information is lost, i.e. they cannot recover the signal voids that are present when using zero-padding reconstruction (as it does not depend on the phase information). Thus, as it sets the bar for all PF methods, it is important to assess the resilience of the zero-padding method against cardiac pulsation artefacts. In this study we investigated this open question and quantified the amount of cardiac pulsation artefacts for various PF fractions. Furthermore, we used FF acquired data to retrospectively create PF data, as well as correct high-resolution (FF) phase images. These were used to study cardiac pulsation artefacts in ideal PF reconstructed images, i.e. using k-space conjugate synthesis with the correct high-resolution phase demodulation.