Heart rate–dependent reduction in right ventricular strain: insights from controlled atrial pacing and fully automated deep-learning echocardiography

Background

Right ventricular (RV) function is crucial for clinical decision-making. However, effects of different physiological conditions on RV free wall longitudinal strain (RV-FWLS), including heart rate variations, are not well characterized. Fully automated deep learning (DL)-based echocardiography now enables fast, reproducible assessment of RV-FWLS, facilitating accurate and efficient evaluation of the influence of heart rate on RV-FWLS.

Purpose

We aimed to quantify how increasing heart rates influence RV-FWLS, using a controlled atrial pacing protocol and fully automated DL-based measurements.

Methods

Participants with implanted programmable pacemakers due to sinus node dysfunction and with preserved atrioventricular (AV) conduction were enrolled. Heart rate was controlled using a fixed atrial pacing protocol, and echocardiographic acquisitions were performed at baseline and at stepwise increments of 10 beats/min up to 140 beats/min. For each heart rate level, 10-beat cine-loops of the RV focused view were acquired. RV-FWLS was measured as the mean of all ten cardiac cycles using a fully automated DL-based method developed and validated by our group [1]. The relationship between heart rate and RV-FWLS was analysed with a linear mixed model, including patient ID as a random effect, and evaluated using a likelihood ratio test.

Results

47 participants were included (53% females), age ranging from 27 to 87 years (mean 68). Heart rate at baseline ranged from 50 to 87 beats/min. Early termination of the protocol was mainly due to development of frequency-related second-degree AV block, resulting in 26 participants reaching 130 beats/min and 16 reaching the target of 140 beats/min. In total, 34 participants were observed through the entire pacing range 80-120 beats/min. RV-FWLS decreased progressively with increasing heart rate, from 24.9% at 60 beats/min to 14.5% at 140 beats/min, representing a 42% relative reduction (95% CI: 25% - 58%), with a mean per-participant absolute decrease in RV-FWLS of 1.3% per 10 beats/min (Figure 1). The strong correlation between HR and RV-FWLS was supported by a highly significant likelihood ratio test (p <0.001). A total of about 3110 cardiac cycles were analysed with a mean processing time of <1sec/cycle.

Conclusions

RV-FWLS decreased linearly with increasing heart rate, highlighting the need to incorporate heart rate when interpreting RV-FWLS in clinical practice. Our findings also illustrate the potential of fully automated DL methods to support high-throughput strain analysis.

[1] Chernyshov A, Grue JF, Nyberg J, Grenne B, Dalen H, Aase SA, Østvik A, Lovstakken L. Automated Segmentation and Quantification of the Right Ventricle in 2-D Echocardiography. Ultrasound Med Biol. 2024;50(4):540-548.Figure 1