Assessment of 3D velocity vector fields and turbulent kinetic energy in a realistic aortic phantom using multi-point variable-density velocity encoding

Background Three-dimensional Phase Contrast (PC) MRI has emerged as a promising non-invasive acquisition technique for assessing velocity vector fields of blood flow [1]. To address the limited sensitivity when velocities are much lower than the encoding velocity venc, three-point acquisition methods with a high venc and a low venc acquisition to unwrap the low venc scan may be employed [2]. However, by using the high venc data only to control phase unaliasing the approaches are not signal-to-noise ratio (SNR) efficient. This fact becomes relevant in particular when incorporating data undersampling techniques to shorten the long scan times associated with 3D PC-MRI. Accordingly, SNR optimality of encoding and decoding is desired. To this end Bayes’ approaches have been proposed and adapted to PC-MRI [3,4]. In the present work the feasibility of velocity vector field and turbulent kinetic energy (TKE) mapping based on multi-point variable-density velocity encoding with spatiotemporal undersampling is demonstrated on a realistic aortic phantom [5]. Methods An elastic cast of an aortic arch equipped with a mechanical aortic valve (St. Jude Medical Inc., St. Paul, MN, USA) was set up in a pulsatile flow conduit and measured using a velocity encoded, cardiac triggered 3D gradient echo sequence on a 3T Philips Achieva System (Philips Healthcare, Best, The Netherlands). Within a scan time of 33 min, 5 velocity encodings according to venc = [200, 100, 50, 28, 20] cm/s in each spatial direction plus a non-encoded reference segment were acquired (Fig 1. red dots) with 5x k-t undersampling and 11x6 training profiles with a temporal resolution of 46 ms. Velocities and TKE values [5] were computed using Bayesian parameter estimation [6]. In a second experiment, one leaflet of the valve was fixed in order to simulate a stenotic valve.