Particle Trace Visualization of Flow Patterns Downstream of a Prosthetic Aortic Valve in Patients S. Kozerke’, J.M. Hasenkam’, E.M. Pedersen2, P. Boesiger’ ‘Inst. of Biomedical Engineering and Medical Informatics, Univ. and ETH, Zurich, Switzerland 21nst. Exp. Clin. Rex and Dept. Cardiothoracic and Vase. Surgery, Aarhus Univ. Hospital, Denmark Introduction Late complications such as hemolysis and thrombus formation in patients with an implanted prosthetic heart valve have been related to fluid dynamic characteristics [ 11. Specific blood flow patterns such as flow reversal and vortex flow have been considered to contribute to thrombus formation [2]. In order to detect alterations of flow fields around prostheses in humans MR velocity mapping might be a powerful modality. Furthermore in-vivo data would help to verify current in-vitro and numerical experiments used for the development of prosthetic valves. Two-dimensional time-resolved velocity distributions around a mechanical heart valve have already been studied with MRI in humans [3]. However the three-dimensional nature of flow downstream of prosthetic valves remains poorly defined as measuredwith two-dimensional methods. To assess three-dimensional flow patterns downstreamof the St. Jude Medical aortic valve in patients particle trace visualization of data obtained with a tine navigator gated 3D phase contrast sequence is presented. Methods Seven patients (3 males, mean age 58.6, range 44-66) with an implanted St. Jude Medical (SJM) aortic were investigated in supine position using a Philips Gyroscan NT 1.5T whole body scanner (gradients: 21 mT/m amplitude, 100 mT/m/ms slew rate). Data acquisition: Time-resolved phase contrast data were acquired using a 3D hybrid sequence(TFEPI) performing k-space segmentation with three excitations each followed by three EPI readouts within each heart phase. The imaging volume was positioned on an early systolic frame obtained with an initial tine scan immediately distal to the artifact arising from the sewing ring of the prosthetic valve. The following parameters were used: FOV 224x157~40 mm3, matrix size 128x90~10, velocity encoding range +lSO cm/s, Te=4.2 ms, Tr=8 ms, trigger delay 62 ms, temporal resolution 35 ms. Velocity encoding was performed in all three spatial directions in subsequent heart beats. For respiratory motion compensation a 2D selective navigator pulse was placed through the right hemi-diaphragm and was applied immediately after detection of the R-wave. A cubic function was used to weight the gating window from 5 mm for the central k,-iines to 12 mm for the most peripheral k,-lines. If within the gating window the offcenter of the imaging volume was corrected according to the shift of the diaphragm position with respect to the end expiratory position. The end expiratory position was automatically defined by means of a prescan. Assuming a heart rate of 70 beats/min and a gating efficiency of 50% scan duration was 8 min. Data processinp: Background phase correction was performed by subtracting a linear function fitted through the phase of stationary tissue for each velocity component. Vessel cross-sections were segmented using active contour based segmentation[4]. Particle trace visualization: A software package was designed to provide particle trace calculation and interactive visualization of particle paths and velocity vector maps in a perspective view. Traces of particles released on an user-defined plane at a desired time within the vessel lumen were generated by time integration using a numerical multi-stage scheme [5] with variable time steps. In space trilinear interpolation of velocity data was used. Interpolation in time was performed by application of cubic polynomials. Since the ascending aorta moves considerably over the cardiac cycle the gravity centers of temporal neighboring vessel cross sections were matched during time interpolation. Vector plots of velocity were used to calculate flow through any plane within the data set. Results Figure la depicts a typical example of paths generated by particles that have been releasedduring systolic flow acceleration (62 ms after R-wave) in a plane perpendicular to the aortic axis 10 mm downstream of the valve (particles were traced for 50 ms, the central slit of the valve is perpendicular to left-right). Two major jets emerging from the lateral orifices of the valve are seen. Releasing particles at the same spatial location later in time (132 ms after R-wave) discloses reverse flow pattern during systole in areasabove the lateral orifices (Figure lb). Figure 1: Particle tracesdownstream of the St. Jude Medical aortic valve. Tracing particles released 62 ms after R-wave show the development of two lateral jets (a). Particles released 132 ms after R- wave disclose reverse flow pattern generateddiring systole (b). Discussion We present the first 3D visualization of flow patterns downstream of the SJM prosthetic valve in humans. The shape of the flow pattern during early systolic flow acceleration reflects the valve design with two distinct lateral jets. Later in systole reverse flow patterns are generated indicating recirculation zones. In contrast to in-vitro findings observed flow patterns appear to be more skewed. This skewness seems to be related to the curvature of the ascending aorta and the angulation of the valve with respect to the aortic and ventricular axis. Animation of particle traces helps to understand complicated flow patterns developed over time. Using this in-vivo information current in-vitro and numerical simulation methods used for prosthetic valve design could be verified. References 1. Woo, Y.R., Yoganathan, A.P. Life Support Syst., 3:283-312; 1985 2. Wurzinger, L.J., Blasberg, P., et al. Biorheology 22:437-50, 1985 3. Houlind, K., Eschen, O., et al. J. Heart Valve Dis. 5:51 I-17, 1996 4. Kozerke, S., Botnar, R., et al. MAGMA Suppl. 4:290, 1996 5. Darmofal, D.L., Haimes R. Compuf. Physics 123:182-195, 1996