Simulation-based investigation of partially parallel imaging with a linear array at high accelerations.

Abstract

Partially parallel imaging strategies such as SMASH, SENSE, and PILS rely on the sensitivity distribution of phased array RF coils to reduce MRI imaging time. Using an N-element phased array, these techniques allow maximum accelerations, L, such that L < or = N, with acceleration defined as the factor by which scan time is reduced in comparison to traditional, fully gradient encoded acquisitions. As N increases in modern MRI facilities or using special hardware extensions, its role as the primary limitation in partially parallel imaging will be reduced and other limiting factors will become dominant. Two such factors include available SNR and the variation of sensitivity distributions with imaging depth. Simulations have been conducted to evaluate the impact of slice depth and noise on partially parallel reconstructions for the case of a square linear array of overlapped elements that are parallel to the imaging plane. Results indicate that even when sensitivity distributions are exactly known, the linear surface array can only provide high accelerations over a limited imaging depth due to changing suitability of the sensitivity distributions for partially parallel reconstruction. This work emphasizes the importance of simulations for target-based partially parallel array design.