Abstract 4146697: Computational Pipeline for Right Ventricular Hemodynamic Analysis from Pressure-Volume Loop Generation

Introduction: In patients with pulmonary hypertension (PH), right ventricular (RV) dysfunction is a key discriminator of clinical outcomes, However, current clinical indices of RV function are poor. The gold standard for evaluating RV function is the invasive pressure-volume (PV) loop, which is not acquired routinely clinically. Moreover, there is no streamlined pipeline for estimating PV loop derived from waveforms obtained by standard chart review. Here, we present a new semi-automated pipeline for single beat estimation of PV loops and load-independent RV contractile indices from clinically obtained waveforms from routine right heart catheterization. Methods: RV pressure waveforms were obtained from PH patients (n=24) ( Figure 1A ). Waveform reports were digitized using computer vision techniques and subsequently averaged for single-beat analysis. P isovol was calculated using a piece-wise sinusoidal fit. End-systolic (ESP) and end-diastolic pressures (EDP) were estimated using the square of the second derivative of the waveform. The end-diastolic pressure-volume relationship (EDPVR) was calculated by fitting an exponential function. Ejection was estimated using a cubic spline between (ESV, ESP) and (EDV, Pressure[dP/dt max ]). Clinically relevant RV functional parameters Ees (end systolic elastance) and Ea (effective arterial elastance) were obtained as previously described. Results: PV loops, Ees, and Ea were derived from 24 patients. Single-beat validation data were acquired by applying single-beat estimation to data collected from a standard pressure catheter. A representative pressure waveform and resulting PV loop estimate are shown in Figures 1A and B , respectively. The Bland-Altman analysis ( Figure 1C ) reported moderate to good agreement between two datasets, with a bias of -0.03 mmHg/mL (95% CI: -0.28, 0.21 mmHg/mL) and 0.08 mmHg/mL (95% CI: -0.29 to 0.45 mmHg/mL) for Ea and Ees, respectively. Conclusion: We have developed a novel and robust computational pipeline tool for digitization and single-beat analysis of RV pressure waveforms from an image of RV pressure waveform. Results from this pipeline were consistent with the single-beat validation data.