The purpose is to validate our cardiovascular image analysis toolbox (called "WaVelocity") for measuring pulse wave velocity (PWV) from cardiac magnetic resonance (CMR) images against PWV measurements of in-place pressure catheter.

The validation was performed using two phantoms: a straight latex tube (length 40cm, diameter 1cm) and an aortic phantom (rubber tube with shape of aorta, length 50cm, diameter 2.5cm). Electric water pump was used to pump water through phantoms with pumping rates of 52 beats/minute and 75 beats/minute. The pump also generated an external gating impulse signal that was used to perform gating for the CMR sequences.

Images were recorded on Siemens Avanto Fit 1.5T scanner as Phase-Contrast (PC) para-sagittal sequences in 3 (for the straight tube) to 5 (for the aortic phantom) planes with velocity encoding (VENC) value 200cm/s for the straight tube phantom and 150cm/s for the aortic phantom. Each sequence (that corresponds to one heart cycle) consists of 40 images. In total, 4 sequences were recorded (for each phantom two pumping rates of 52bpm and 75bpm were used).

The same phantom setting was used to perform pressure measurements with a pressure catheter. The catheter was initially placed at the beginning of the tube (corresponding to the ascending aortic level for the aortic phantom) and pressures were recorded during the pullback of the catheter while sampling the pressure every 5cm along the centerline.

Phase-Contrast images were processed as follows. The centerline of the tubular phantom was extracted using a variation of our vessel centerline extraction method¹, after which the position markers were placed equidistantly at 5 cm apart along the centerline as in Figure 1. For each of the position markers the average velocity was is calculated as average grey value of pixels belonging to the region of the phantom in the predefined neighborhood. This results in velocity curves for each of the positions with 40 time samples (see Figure 2). In order to determine the time shifts between the velocity curves, we used either the foot-of-the-curve or the cross-correlation method (if the foot of the curve was clearly visible the corresponding method was used, otherwise we used the cross-correlation). From the measured lengths and time shifts between different position markers the pulse wave velocity was calculated².

The time shifts between the pressure curves obtained by catheterization at equidistant positions along the centerline were calculated using foot-of-the-curve method.

Results for PWV measurements for the first marked position (corresponding to ascending aortic level) as the reference position are shown in Figure 3 for the tube phantom and Figure 4 for the aortic phantom. Mean and error values are as follows. For the tube phantom at 52bpm: MR mean PWV 2.37m/s, Catheter PWV 1.89m/s, relative error 0.3, absolute error 0.48m/s; tube phantom at 75bpm: MR mean PWV 2.07m/s, Catheter PWV 2.45m/s, relative error 0.15, absolute error 0.38m/s; aortic phantom at 52bpm: MR mean PWV 2.47m/s, Catheter PWV 2.04m/s, relative error 0.21, absolute error 0.42m/s; aortic phantom at 75bpm: MR mean PWV 2.29m/s, Catheter PWV 3.63m/s, relative error 0.37, absolute error 1.34m/s.

The results show sufficiently high correspondence between the calculated MR PWV values and the PWV values measured using the pressure catheter. Absolute errors will not have significant diagnostic or prognostic relevance (except for aortic phantom at 75bpm, which is explained later in this section) since PWV values for normal population fall in range [2,10]m/s. Velocity curves in the aortic arch (for the aortic phantom) do not correspond to real velocity curves due to the fact that the direction of the water flow in the arch is parallel to the axis of velocity encoding. This does not pose a real problem because the PWV measurement for the aorta is most often performed at the most critical levels (ascending, descending, diaphragmal and abdominal) for which this problem does not arise. Consequently, we also skip PWV calculations in the aortic arch. Initially slight irregularities were detected in MR PWV measurements for aortic phantom at 75 beats/minute. However, further inspection of CMR images showed that the PWV values were accurately calculated, and that the irregularities are the result of reflective waves.Our cardiovascular image analysis toolbox "WaVelocity" has proven accurate for measuring pulse wave velocity (PWV) from cardiac magnetic resonance (CMR) images when compared with PWV measurements of in-place pressure catheter. The future work will incorporate validation against gold-standard pressure catheter measurements in patients.