Precision cardiac monitoring: algorithmic real‐time assessment of right ventricular function in pulmonary hypertension

Right ventricular (RV) dysfunction significantly impacts both mortality and symptomatology in patients with pulmonary hypertension. To assess RV function, metrics such as RV‐arterial coupling (calculated by dividing end‐systolic elastance by arterial elastance) or, more simply, the RV ejection fraction (RVEF) are used. ¹ Although cardiac magnetic resonance imaging (cMRI) is the established benchmark for RVEF quantification, it lacks suitability for real‐time monitoring of acute changes in response to therapeutic interventions. Thus, it falls short in tracking treatment effectiveness or disease progression at the bedside. Recently, Heerdt et al. introduced and validated an innovative approach that derives RVEF exclusively from the RV pressure signal using the following equation:

$$\text{RVEF} \text{=} 1‐\mathrm{ESP}/{\mathrm{P}}_{\max }$$ (1)

where P_max is the theoretical maximum isovolumetric pressure of the right ventricle and ESP is end‐systolic pressure. ² , ³ To transition this conceptual advancement to practical bedside application with real‐time calculation, simplification of the detection of P_max and ESP is imperative. Hence, we developed an algorithm for P_max and ESP calculation, ensuring reproducibility and consistency in app‐based RVEF calculation.

Our study complies with the Declaration of Helsinki, and the local ethics committee approved both parts of the study [Approval No. 108/15 (retrospective part) and Approval No. 211/22 (prospective part)]. All participating patients provided informed consent. The proposed algorithm reduces complexity by requiring only three points on the RV pressure curve or its first derivative to estimate the RVEF: systolic pressure and the maximum and minimum of the first derivative of pressure with respect to time (dP/dt_max and dP/dt_min, respectively). For P_max calculation, the algorithm identifies the isovolumetric contraction point (dP/dt_max) and relaxation point (dP/dt_min) on the RV pressure curve. These serve as initial points for two tangential lines, where the line slope is set as dP/dt_max for the ascending segment and dP/dt_min for the descending segment (Figure 1 A ). Utilizing these slopes, the starting pressure (pressure at dP/dt_max and dP/dt_min), and ejection time [ejt; defined as the time difference between the two points dP/dt_max and dP/dt_min (Equation 2)], we calculate the tangents and their intersection point [P_cross (Equation 4)], whose y‐value is converted to P_max via the equation of Shih et al. ⁴ (Equation 5) using mean pulmonary artery diastolic pressure (PADP) as shown below:

$$\text{ejt} \text{=} t@\text{dP}/{\mathrm{dt}}_{\min }‐t@dP/{\mathrm{dt}}_{\max }$$ (2)

$$\mathrm{t}1=\left(\right(\mathrm{P}@\mathrm{dP}/{\mathrm{dt}}_{\min }-\mathrm{P}@\mathrm{dP}/{\mathrm{dt}}_{\max })-\mathrm{dP}/{\mathrm{dt}}_{\min }\times \mathrm{ejt})/(\mathrm{dP}/{\mathrm{dt}}_{\max }-\mathrm{dP}/{\mathrm{dt}}_{\min })$$ (3)

$${\mathrm{P}}_{\text{cross}}=P@dP/{\mathrm{dt}}_{\max }+\mathrm{dP}/{\mathrm{dt}}_{\max }\times \text{t1}$$ (4)

$${\mathrm{P}}_{\max }=2/{3.141}^{*}\left({\mathrm{P}}_{\text{cross}}-\text{mean}\mathrm{pad}\right)+\text{mean}\mathrm{pad}$$ (5)

$$\mathrm{mean} \mathrm{pad}=\left(\text{P@dP}/{\mathrm{dt}}_{\max }+P@dP/{\mathrm{dt}}_{\min }\right)/2$$ (6)

Figure 1.

(A) Modified P_max calculation. (B–D) Bland–Altman analysis of (B) automated calculation of P_max and conventional P_max, (C) eESP and 2dESP, and (D) automated calculation of RVEF and cMRI‐derived RVEF in a retrospective analysis of 99 patients with and without pulmonary hypertension. (E–G) Comparison of automated calculation of RVEF with 3D echocardiographic RVEF in a prospective analysis of five patients with pulmonary hypertension undergoing continuous measurement of RV pressure via the CorLog Probe; (E, F) longitudinal data from two example patients and (G) Bland–Altman analysis of data from all five patients (excluding Day 0 when the calibration factor was determined) are shown. Echo, echocardiographic; ejt, ejection time; P, pressure; P_max‐auto, automated calculation of P_max; P_max‐fit, conventional P_max; t, time.

Our method simplifies P_max calculation, requiring only a single point and its slope, unlike previous methods that used multiple points during isovolumetric phases. It should be noted that the original method of Shih et al. ⁴ was described for the left ventricle and has not yet been investigated in the right ventricle. Additionally, the method used by Heerdt et al. relied on estimation of P_max using a four‐parameter Weibull peak fit, which showed some agreement with the sinusoid fitting technique. ²

The state‐of‐the‐art technique to determine ESP is detection of the pressure at the time of the initial minimum of the second derivative of pressure over time (d²P/dt²) before the dP/dt minimum. ⁵ Nonetheless, due to signal smoothening (essential for dependable pressure point detection), the second derivative frequently lacks a detectable minimum. Consequently, the algorithm computes an estimated ESP (eESP) by averaging the systolic and minimum dP/dt pressure detection points. The algorithm subsequently employs Equation (1) to perform real‐time calculations of RVEF on a beat‐to‐beat basis.

In a first step, we validated our algorithm using retrospective data from 99 patients with (n = 92) and without (n = 7) pulmonary hypertension who had undergone RV catheterization and cMRI within 24 h as part of the Right Heart I study (ClinicalTrials.gov Identifier: NCT03403868).

The automated method for estimating P_max demonstrated a strong correlation with the conventional sinusoid fitting method ⁵ (r = 0.97, P < 0.001). However, a Bland–Altman analysis indicated a mean discrepancy of 8.2 mmHg and limits of agreement from −5.0 to +21.3 mmHg, suggesting a minor tendency of the automated method to underestimate P_max (Figure 1 B ). Similarly, eESP correlated well with manual detection from the second derivative of the ventricular pressure signal (2dESP; r = 0.98, P < 0.001), with a mean difference of 5.4 mmHg and limits of agreement between −5.2 and +16.0 mmHg (Figure 1 C ). This analysis also indicated a trend for the automated method to slightly underestimate ESP, particularly at higher values, which may reflect the intrinsic challenge of empirically determining ESP as it approaches peak systolic pressure in conditions of elevated systolic pressures. ⁶ The comparison between calculated and cMRI‐derived RVEF demonstrated a strong positive correlation (r = 0.81, P < 0.001), with Bland–Altman analysis revealing a minimal mean bias of 1% and limits of agreement spanning from −16% to +17% (Figure 1 D ). Analogous to ESP, the Bland–Altman distribution for RVEF was non‐symmetrical, indicating that the algorithm tends to overestimate RVEF in patients with lower RVEF values and underestimate it in those with higher RVEF values. Reasons for the observed distribution include the tendency of the automated P_max method to underestimate P_max compared with the sinusoid fitting technique and the underestimation of ESP at high amplitudes and overestimation at lower amplitudes by eESP. Furthermore, tricuspid insufficiency, which is common in pulmonary hypertension, can alter the isovolumetric pressure rise, impacting the accuracy of P_max prediction and, consequently, RVEF estimation.

To assess the algorithm's proficiency in monitoring patients' longitudinal EF changes—a crucial capability for a monitoring algorithm—we employed our algorithm to calculate RVEF by analysing pressure tracings obtained through the CorLog device (emka medical GmbH, Germany). ⁷ This prospective analysis was conducted in a cohort of five hospitalized patients who had been diagnosed with pulmonary hypertension. The algorithmic approach was then compared with RVEF values obtained through 3D echocardiography. Upon calibration against the RVEF derived from 3D echocardiography on the day of CorLog Probe implantation, the algorithm demonstrated a remarkable capacity to track precisely subsequent echocardiographic RVEF measurements over the ensuing days (as illustrated in Figure 1 E,F ). The calibration procedure entailed quantifying the disparity between 3D echocardiographic and CorLog RVEF measurements on Day 0 and subsequently fine‐tuning the algorithm by this computed margin. Bland–Altman analysis revealed a minimal bias and limits of agreement ranging from −6% to +6% (Figure 1 G ).

Drawing from previous work, this algorithm presents a single‐beat pressure methodology that streamlines and automates the computation of P_max and ESP. This facilitates the real‐time beat‐to‐beat assessment of these parameters, for example, through an app interface, thereby enabling a continuous online evaluation of RVEF. Notably, despite the algorithm's streamlined approach, it showed a good level of accuracy when compared with the traditional offline manual sinusoid fitting technique in our comprehensive investigation involving 99 patients. We believe the merit of pressure‐based RVEF assessment lies primarily in longitudinal monitoring, where algorithm accuracy and limits of agreement can be enhanced through calibration to a known volume rather than in the exact determination of RVEF a priori. Although this algorithm's pressure‐based estimation of RVEF cannot match the precision of cMRI, it provides a valuable bedside tool for clinicians, offering insights into RV performance for informed decision‐making in managing complex conditions like heart failure and severe pulmonary hypertension.

Funding

This work was funded by the Excellence Cluster Cardio‐Pulmonary System and the Collaborative Research Center (SFB) 1213—Pulmonary Hypertension and Cor Pulmonale, grant number SFB1213/1, project B08 (German Research Foundation, Bonn, Germany).

Kremer, N. , Glocker, F. , Schäfer, S. , Rako, Z. , Yogeswaran, A. , Seeger, W. , Hopf, H.‐B. , and Tello, K. (2024) Precision cardiac monitoring: algorithmic real‐time assessment of right ventricular function in pulmonary hypertension. ESC Heart Failure, 11: 2469–2472. 10.1002/ehf2.14833.