Smartwatch Signals Could Reveal Hidden Heart Failure Through the Beat After a Skipped Pulse

Every so often, the human heart fires an early, extra beat. These premature ventricular contractions, or PVCs, are so common that most people experience them without ever noticing, and in otherwise healthy individuals they are usually harmless. But the beat that follows a PVC is anything but ordinary. After the premature contraction and the compensatory pause that trails it, the heart muscle contracts with noticeably more force than usual, a phenomenon known as post-extrasystolic potentiation, or PESP. Cardiologists have studied this boost in contractility for decades, and they have long known that its strength and its decay differ markedly between healthy hearts and failing ones. A team of researchers at Eindhoven University of Technology, Catharina Hospital Eindhoven, and Philips Research now reports a way to read that difference from a signal as ordinary as the one produced by a smartwatch, potentially turning an everyday wearable into a continuous monitor of cardiac function.

The study, published in Frontiers in Signal Processing, hinges on photoplethysmography, or PPG, the optical technique that underlies the heart-rate sensors in most consumer wearables. A PPG sensor pairs a light-emitting diode with a photodiode placed against the skin, illuminating the tissue beneath and measuring the changing volume of blood flowing through it with each heartbeat. Because the hardware is cheap, non-invasive, and unobtrusive, PPG has become ubiquitous in smartwatches, rings, and finger clips. It reliably yields heart rate and heart rate variability, and it can even be measured remotely from video recordings. Yet the technique has well-known weaknesses. Wrist-based recordings are easily contaminated by motion artifacts, and the signal lacks a true physical unit: it reflects relative changes in blood volume rather than absolute quantities, and the amount of light reaching the photodiode depends on skin pigmentation, sensor contact, posture, peripheral vascular resistance, blood pressure, and cardiac output. For that reason, most PPG applications rely on relative measures such as amplitude ratios rather than absolute values.

The physiology behind PESP begins with the premature beat itself. A PVC typically has a shortened filling time and ejects little blood, because the cardiac muscle cells have not fully recovered from the previous contraction and venous return during the abbreviated diastole is limited. The compensatory pause that follows allows the ventricle to fill more completely, and the next beat arrives with augmented contractility and output. Part of this augmentation can be explained by the Bowditch effect, in which cardiac inotropy depends on heart rate, and by the Frank-Starling principle, in which stroke volume depends on left ventricular end-diastolic pressure. But PESP behaves as a phenomenon in its own right, acting separately from these classical mechanisms. PVCs are common in the general population and become more frequent with age, underlying heart disease, and comorbidities such as hypertension or coronary artery disease. Crucially, a higher occurrence of PVCs and an enhanced PESP are typical findings in heart failure.

After the potentiated beat, the extra contractility fades over the next few heartbeats, returning to baseline within several cycles. Earlier work on human cardiac tissue showed that this decay follows an inverse exponential function with a time constant of roughly 800 milliseconds, and that in heart failure the mechanical restitution is slower, corresponding to a larger time constant. More striking still, researchers have observed oscillations in the maximum arterial blood pressure following a PVC in patients with severe heart failure. This oscillating behavior is attributed to disrupted calcium cycling, specifically a sarcoplasmic reticulum depleted of calcium through leaky ryanodine receptors and reduced calcium uptake, partly compensated by upregulation of the sodium-calcium exchanger in the plasma membrane. Animal studies have reproduced the same oscillatory signature. Until now, however, observing these oscillations required invasive blood pressure measurement, which rules out continuous monitoring in daily life.

The Dutch team hypothesized that the oscillating pressure behavior seen in severe heart failure should also leave a trace in the amplitude of PPG signals, which are influenced by cardiac contractility, cardiac output, and blood pressure. Their approach models the decay of potentiation after a PVC as a second-order system, the same mathematical framework that describes a mass-spring-damper mechanism or an RLC electrical circuit. Such a system can capture both the gradual increase associated with mechanical restitution and the oscillatory response to an extrasystolic beat seen in severe heart failure. The researchers then reduced the model to a single meaningful parameter, the damping factor, and investigated how it relates to the ejection fraction, the standard clinical measure of how much blood the heart expels with each contraction, in patients diagnosed with heart failure with reduced ejection fraction.

The data came from an internal dataset collected during percutaneous electrophysiological ablation procedures, in which the heart is paced at a fixed interbeat interval. The fixed pacing keeps cardiac output constant, minimizing the confounding influences of the Bowditch and Frank-Starling effects, before a PVC is deliberately introduced and pacing is paused to let the heart respond with its compensatory pause. Electrocardiogram and arterial blood pressure were recorded at 1000 Hz with a GE Mac-Lab Hemodynamic Recording System, while PPG was captured at 32 Hz by a Philips Elan device worn on the wrist like a watch. Ejection fractions were obtained by echocardiography up to six months before the procedure, as clinical routine, with the patients clinically stable so that no significant change in ejection fraction was expected. The dataset comprised seven patients, all male, with ejection fractions ranging from 24 to 56 percent, contributing a total of 394 PVCs whose post-extrasystolic responses were undisturbed by large artifacts.

Signal processing began with bandpass filtering of the raw PPG between 0.07 and 12 Hz to suppress baseline wander and high-frequency noise, and low-pass filtering of the arterial blood pressure above 12 Hz. Pulses were detected with a peak detector, and each pulse was defined as the signal between consecutive systolic upslopes. An expert annotator, guided by the electrocardiogram where the large electrical activation of a PVC is easily recognized, labeled each pulse as a PVC pulse, a potentiated pulse, or a normal pulse. Only PVCs preceded by a normal pulse and followed by five non-PVC pulses were included, ensuring that the modeled recovery was uncontaminated. For each event, the pulse amplitude was defined as the absolute difference between the systolic peak and the start-of-diastole value, and the sequence of amplitudes following the PVC was fitted with the second-order model, treating the PVC as a Dirac delta input with an amplitude offset. Model parameters were estimated by minimizing the mean squared error using the limited-memory BFGS algorithm in Python, first per event and then across all of a patient’s events to handle large amplitude variations caused by posture changes or wristband tightness.

Two simplifications made the model practical. First, following earlier suggestions, the time axis was normalized by beat number rather than by interbeat interval, which fixes the natural frequency of the system and reduces the number of free parameters. Second, because oscillations in pulse amplitudes occur between consecutive pulses, the input was shifted by half a beat so that the model’s extreme values coincide with the pulse peaks. The result is an optimization problem in which only the damping factor describes the behavior of the PESP decay between patients. When the system is overdamped, the response rises quickly but falls short of its full gain-determined value and decays exponentially back to baseline; increasing the damping slows the rise and delays the decay. When the system is underdamped, the response overshoots and oscillates around the baseline with decaying amplitude, and lower damping produces stronger, longer-lasting oscillations. To guard against artifacts and outliers, the researchers ran the entire procedure 256 times per patient on randomly sampled subsets containing 70 percent of the PESP responses, a bootstrapping strategy that yields a distribution of damping values for each patient.

The results revealed a clear pattern. For both the invasive arterial blood pressure signal and the non-invasive PPG signal, the damping factor increased as ejection fraction decreased, until a critical point at an ejection fraction of roughly 30 percent, which the authors call the collapsing point. Below that threshold, the damping factor dropped below one, the mathematical signature of an underdamped system, indicating persistent oscillations in the pulse amplitudes. The variance of the damping estimates was larger for PPG than for arterial blood pressure, as expected given PPG’s lower signal-to-noise ratio and greater susceptibility to confounders, but the same trend and the same collapsing point remained visible. The findings echo earlier work by Sinnecker and colleagues, who found that oscillations in systolic blood pressure after a PVC appeared only in the non-survivor group of their cardiac cohort, suggesting that such oscillations may indicate severe heart failure.

The authors are careful to frame the study as exploratory. With only seven patients, strong conclusions cannot be drawn, and even 394 PVCs cannot be guaranteed to be fully independent events. The method also currently depends on PVC-induced disturbances and an otherwise normal heart rhythm, and PVC detection in PPG signals alone remains non-trivial, since the reduced blood expelled during a PVC can make it appear merely as an extended diastole of the preceding pulse. The data were recorded in a controlled clinical environment, and everyday motion artifacts may complicate real-world application. Still, the correspondence between damping values derived from invasive blood pressure and from a wrist-worn optical sensor suggests the phenomenon is robust to measurement modality, because the model’s scaling factor absorbs amplitude differences. If validated in larger cohorts, including datasets such as MIMIC-III, the approach could turn the smartwatch already on millions of wrists into an early-warning system for declining cardiac contractility, shrinking the interval between a silent deterioration and a hospital admission.

Subject of Research: Non-invasive assessment of post-extrasystolic potentiation via PPG to gauge cardiac contractility and heart failure severity

Article Title: Modeling the post-extrasystolic potentiation to determine heart failure

Article References: Montree, R. J. H., Peri, E., Dekker, L. R. C., Long, X., Haakma, R., & Vullings, R. (2026). Modeling the post-extrasystolic potentiation to determine heart failure. Frontiers in Signal Processing, 6, Article 1909928. https://doi.org/10.3389/frsip.2026.1909928

Image Credits: AI Generated

DOI: 10.3389/frsip.2026.1909928

Keywords: heart failure, post-extrasystolic potentiation, photoplethysmography, premature ventricular contractions, ejection fraction, damping factor, second-order system, wearable devices, cardiac contractility, biomedical signal processing, smartwatch, calcium cycling