A novel semi-automated method to improve estimation of right ventricular systolic pressure by Doppler ultrasound

Abstract

Accurate measurements of maximal tricuspid regurgitation (TR) velocity is central for noninvasive estimation of right ventricular systolic pressure (RVSP). A better understanding of TR waveform shape may improve confidence in RVSP estimation. Our aim was to develop a semi-automated analysis of TR waveform shape and to model TR Vmax. We created an automated method to extract TR signals from the doppler image. We used Spearman correlation to study features of TR waveforms in relation to RV metrics. We then used a multi linear regression to develop a model to predict RVSP. From a cohort of 115 patients with pulmonary hypertension, we randomly defined a derivation (67%) and a validation data set (33%). We achieved good extraction of the waveform from the doppler (Pmax and VTI extracted vs. Pmax and VTI measured, r 2 = 0.991 and 0.929). Curve characteristics such as Pmax, skewness, defined as time to Pmax/RR, dP/dt max and min are strongly associated with RV longitudinal strain ( Table 1 ). The multivariate analysis using dP/dt min and max, pressure at dp/dt min and max time points, heart rate and time of TR duration normalized by RR predicted RVSP in our training data set (r 2 = 0.908). The model enabled prediction of RVSP in the validation data set (r 2 = 0.949, Fig. 1 ). We have developed an automated model to analyze TR waveforms and showed that interpolation of TR waveform features is possible and can provide additional support for accurate estimation of RVSP.