P2403The development of automated methods for the reproducible assessment of aortic stenosis

Abstract

Background: Aside from symptoms, the principal determinant of surgical timing in aortic stenosis is echocardiography. Our previous work found that the dominant sources of variation were not differences between operators or readers, but, differences in probe placement, beat-to-beat variability, and tracing. This knowledge allowed the development of a novel automated algorithm, the “instantaneous Dimensionless Index (iDI)” that reduces the influence of these factors to improve precision. Methods: Vendor-independent, open-source software for processing Doppler tracings both in “real-time” and retrospectively was developed (Figure 1). Runs of multiple beats are acquired using continuous wave (A). Automatic thresholding, quality control (B), tracing and extraction are performed before overlaying and combining the multiple-beats (C). From this single ensemble beat (D) the outer aortic and inner LVOT tracing are simultaneously extracted (E) to produce the iDI. Thirty patients with aortic stenosis were prospectively recruited. We obtained multiple Doppler recordings using both the traditional peak velocity dimensionless index (DI) and our new method in each patient. Results: The mean coefficient of variation was substantially smaller with iDI as compared to DI (5.5% vs 12.1%, p<0.001). The mean DI was however lower than the iDI (-0.04, p=0.01), often due to underestimation of LVOT and over-estimation of the aortic velocity using the traditional technique. Conclusions: The reduction or elimination of beat-to-beat, tracing, probe-placement, and reader variability lead to a halving of the coefficient of variation. This doubling of precision permits trials using valve severity as an endpoint to require four times fewer patients. In routine clinical clinicians would have greater confidence in measurements and patients would require fewer visits.