Pulse wave velocity (PWV) is a commonly used surrogate for arterial stiffness. One technique that has been used in the past to estimate PWV from MRI or ultrasound velocity data is the foot-to-foot method (1, 2). However, studies have shown that the estimation of PWV by this method can be affected by reflections and by choice of fiduciary point for the “foot” of the flow waveform (3). Here we describe a new method to estimate arterial PWV by using MRI phase-contrast data to tune a 1D blood flow model describing the hemodynamics and propagation of the arterial pulse wave (Fig. 1). Unlike the foot-to-foot method, the proposed method does not depend on choice of fiduciary point. Also, unlike the foot-to-foot method, it accounts for reflections from both within and outside the computational domain.Phase-contrast MRI data were acquired in a single breath-hold from an oblique sagittal plane containing a long section of descending aorta, with velocity encoded in the cranio-caudal direction, and with a velocity-encoding strength (VENC) of 150 cm/s and a temporal resolution of 24 ms. The MRI magnitude images were semi-automatically segmented to obtain the vessel boundaries (yellow lines, Fig. 1) and centerline at each time frame. This was done by seeding 10-15 points along either side of the aorta, fitting each point to the edges using an error function at each phase of the cardiac cycle, and using spline interpolation to determine edges at intermediate locations. The flow rate at a given time and position along the aorta was calculated from the vessel diameter and the spatial velocities measured across the width of the vessel at that location. A fairly straight segment of the vessel of length ~10 cm was analyzed in the current study. The vessel centerline, cross-sectional areas and flow rates determined from MRI data were used to construct a 1D blood flow model of the vessel segment. The flow rate computed at the proximal end of the vessel segment was used to impose inflow boundary conditions on the 1D model while a 0D lumped model was used as outflow boundary condition. The unknowns in the model, vessel and lumped model compliance, were determined by minimizing the difference between predicted flow rates and measured flow rates at three evenly spaced locations other than the inflow boundary. PWVs along the vessel segment were computed once vessel compliance was known. To simulate the lower temporal resolution seen in 4D Flow MRI, we down sampled the original data set to 48 ms and 72 ms, re-computed the flow rates and cross-sectional areas and re-ran calculations with the 1D model. Further, we compared our results against the foot-to-foot tracking method for both the baseline and down-sampled cases using flow rates computed at 4 locations, inlet and the three locations used for model tuning. The fiduciary point chosen for the foot was 15% of the peak flow.Figure 2 shows the PWVs computed for the baseline and down-sampled cases for both the present approach as well as the foot-to-foot tracking method. While the variability in the computed PWVs with the present approach increases slightly as the temporal resolution is degraded, it remains much lower than that obtained with the foot-to-foot method.Results of the current study indicate that the proposed approach could potentially be used with low-time-resolution methods such as 4D Flow MRI to obtain PWV in the aorta with much lower variability than the foot-to-foot method. Further studies are planned to validate the PWV values obtained with the present method with high resolution PC-MRI.