An Efficient Correction Technique for Constant, Linear and ‘Oblique’ Phase Errors in EPI-PROPELLER

Introduction Short-axis and long-axis PROPELLER echo planar imaging (referred to as SAP-EPI and LAP-EPI, respectively) [1, 2] have been developed as alternatives to fast spin echo (FSE)-based PROPELLER sequences [3]. EPI-PROPELLER typically requires fewer shots than FSE-PROPELLER, and is less SAR-intensive. Like other EPI sequences, however, both SAPand LAP-EPI are sensitive to eddy currents that lead to Nyquist ghosts, therefore requiring phase corrections. Conventional phase-correction techniques rely on reference scans [4]. Since each PROPELLER blade activates a different combination of the physical gradient axes, reference scans on a per-blade basis are typically needed, especially for a gradient system exhibiting non-negligible degree of gradient anisotropy [5-8]. Blade-specific reference scans can considerably increase the total scan time of the otherwise very fast SAPor LAP-EPI techniques. In this study, we have developed a time-efficient reference scan method to address phase errors in EPI-PROPELLER sequences. In addition to correcting the constant and linear phase errors common to all EPI sequences, our technique is also capable of reducing the phase error along the phase-encoding direction, a phenomenon responsible for the so-called oblique Nyquist ghost (ONG) [6-8] that is particularly relevant to EPI-PROPELLER sequences. Methods Without losing generality, we demonstrate the proposed technique using a 2D axial scan (in the x-y plane) as an example. The technique acquires only two reference scans along each of the two physical axes, x and y, respectively. The constant (c) and linear (l) phase errors are calculated from each reference scan and denoted by c|| and l|| for the x-axis and c┴ and l┴ for the y-axis (Figs. 1a and 1b). For any arbitrary blade orientation θ shown in Fig.1, the constant and linear phase errors (cθ and lθ, respectively) are calculated using Eqs. (1) and (2) which can be derived with the aid of Figs. 1a and 1b, respectively. With the phase errors now known, phase correction can proceed as if a blade-specific reference scan were acquired. The inconsistent k-space shift along the phase-encoding direction, ' pe k θ Δ , which originates from gradient anisotropy [5-8] in oblique blades (i.e., θ ≠ 0° or