# The impact of heart rate on echocardiographic measures of left ventricular function: novel insights facilitated by deep learning

## TL;DR

This study shows that heart rate significantly affects echocardiographic measurements of the left ventricle, using deep learning to automate and speed up the analysis.

## Contribution

The study introduces a deep learning-based method for fully automated, beat-to-beat echocardiographic analysis to explore heart rate effects on LV function.

## Key findings

- Heart rate increases caused significant, near-linear reductions in LV global longitudinal strain, ejection fraction, and volumes.
- Deep learning enabled rapid and accurate analysis of over 10,000 cardiac cycles with 97% feasibility.
- Findings highlight the need to consider heart rate when interpreting echocardiographic measurements in clinical settings.

## Abstract

Echocardiographic measurements of the left ventricle (LV) are fundamental in diagnosing and monitoring cardiac disease. Still, current understanding of how heart rate influences these measurements is incomplete. We aimed to explore the relationship between heart rate and LV global longitudinal strain (GLS), ejection fraction (LVEF), end-diastolic (LVEDV), and end-systolic volumes (LVESV), using atrial pacing and a transparent multi-step deep learning (DL)-based method for fully automated measurements.

Fifty participants with permanent pacemakers were enrolled. Heart rate was increased by atrial pacing in increments of 10 beats/min, from 50 to 140 beats/min, with echocardiographic 10-beat cine-loops recorded at each step. A DL-based method was utilized to measure GLS, LVEF, LVEDV, and LVESV at all levels.

A total of 10 161 heart cycles were analysed, with 97% feasibility. As heart rate increased, all LV measures displayed significant and near-linear reductions. From 60 to 140 beats/min, GLS decreased by 32% (95% CI: 19–44%), LVEF by 33% (95% CI: 19–47%), LVEDV by 31% (95% CI: 19–43%), and LVESV by 10% (95% CI: −5% to 24%). Processing time per cardiac cycle was 1.3 (0.4) s, corresponding to 3.7 h for the entire dataset.

Heart rate significantly influences echocardiographic measures of LV function and volume, emphasizing the necessity of incorporating heart rate into clinical interpretation and reporting of echocardiographic measurements. This study further demonstrates the potential of DL to advance cardiovascular research by enabling rapid, accurate, and reproducible analyses, previously unachievable due to the inherent constraints of manual measurements.

Central IllustrationThis study aimed to explore the effect of heart rate on common echocardiographic measurements of left ventricular function using atrial pacing and a deep learning-based method. Deep learning enabled fully automated beat-to-beat echocardiographic analysis of 10 161 cardiac cycles. Findings reveal significant heart rate–dependent changes in key LV measures, emphasizing the need to account for HR in clinical echocardiography. A4C, apical four-chamber; A2C, apical two-chamber; APLAX, apical long axis; IQR, interquartile range; LVEF, left ventricular ejection fraction; bpm, beats per minute; GLS, global longitudinal strain.

This study aimed to explore the effect of heart rate on common echocardiographic measurements of left ventricular function using atrial pacing and a deep learning-based method. Deep learning enabled fully automated beat-to-beat echocardiographic analysis of 10 161 cardiac cycles. Findings reveal significant heart rate–dependent changes in key LV measures, emphasizing the need to account for HR in clinical echocardiography. A4C, apical four-chamber; A2C, apical two-chamber; APLAX, apical long axis; IQR, interquartile range; LVEF, left ventricular ejection fraction; bpm, beats per minute; GLS, global longitudinal strain.