Pulse oximeters do more than measure oxygen levels. The pulsing waveform they display, called a photoplethysmogram or PPG, carries a surprisingly detailed fingerprint of how your heart is beating. An abnormal waveform can reveal irregular rhythms such as atrial fibrillation and premature beats, often before you notice any symptoms. The tricky part is that not every jagged or flattened waveform signals a heart problem; movement, poor circulation, and even skin pigmentation can distort the signal in ways that mimic arrhythmias.
What a Normal Pulse Oximeter Waveform Tells You
A healthy PPG waveform has a characteristic shape: a sharp upstroke as the heart ejects blood, a rounded peak, then a gradual downslope interrupted by a small notch called the dicrotic notch. That notch corresponds to the moment the aortic valve snaps shut, and its presence is a sign that your arteries are reasonably elastic and your cardiovascular system is behaving normally. When researchers studied nearly 170,000 participants, they found that an absent dicrotic notch on finger PPG was associated with older age and existing cardiovascular disease.1PubMed Central. Machine Learning to Understand Genetic and Clinical Factors Associated With the Pulse Waveform Dicrotic Notch So even the subtlest features of a “normal-looking” waveform carry clinical meaning.
The height of the waveform, often expressed as the perfusion index, reflects how much blood is pulsing through the tissue under the sensor. A taller, fuller wave generally indicates good peripheral blood flow. A stiffer artery reduces the amount the vessel can expand with each heartbeat, which shrinks the waveform’s amplitude, narrows its width, and flattens the area under the curve.2Scientific Reports. Arterial stiffness assessment using PPG feature extraction and significance testing in an in vitro cardiovascular system Related measures like arterial tension time, which captures how long the pulse wave spends in its upstroke, have also been linked to subclinical atherosclerosis and reduced stroke volume.3PubMed. Arterial tension time reflects subclinical atherosclerosis, arterial stiffness and stroke volume All of this means that when you look at a pulse oximeter screen, you are seeing a miniature portrait of both the heart’s rhythm and the vascular system’s health.
How Irregular Heartbeats Distort the Waveform
When your heart beats in a steady rhythm, every PPG pulse looks nearly identical, spacing out at regular intervals like a row of evenly placed hills. An irregular heartbeat disrupts that pattern in predictable ways. A beat that comes too early, such as a premature ventricular contraction (PVC), produces a pulse wave that is shorter and often weaker because the heart had less time to fill with blood. The beat that follows is usually delayed and stronger, creating an alternating short-tall pattern on the screen. Atrial fibrillation, by contrast, produces beat-to-beat intervals that look essentially random, with no two gaps quite the same and pulse heights that vary erratically.
In some cases the heartbeat is so premature or so weak that it fails to produce a detectable pulse at the fingertip at all. This phenomenon, known as the pulse deficit, is especially common in atrial fibrillation. A recent study used wavelet transformation to make these “missing” pulses more visible in PPG data and then trained a neural network to classify atrial fibrillation based on the pulse deficit pattern. The approach achieved an average area under the curve of 0.975 and an F1 score of 0.935, confirming that the pulse deficit itself is a robust marker of the arrhythmia.4PubMed. Pulse Deficit in Photoplethysmography as an Indicator for Atrial Fibrillation
Detecting Atrial Fibrillation From the Waveform
Atrial fibrillation (AF) is the irregular rhythm that has received the most research attention in the PPG world, partly because it is common, partly because it raises stroke risk, and partly because its chaotic timing leaves a very distinctive stamp on the waveform. Rather than measuring electrical signals the way an ECG does, the oximeter picks up irregular pulse intervals and amplitude variability that serve as indirect but surprisingly reliable surrogates.
Deep learning models have pushed detection accuracy into territory that would have seemed far-fetched a decade ago. One study compared traditional statistical analysis of heart rate variability against neural networks fed raw PPG data. Standard logistic regression on heart rate variability produced an area under the curve of about 0.72. A long short-term memory network using heart rate data improved that to 0.95. And a deep neural network trained directly on the raw waveform reached an area under the curve of 0.98, with sensitivity above 98%.5Heart Rhythm O2. Atrial fibrillation detection from raw photoplethysmography waveforms: A deep learning application A separate hybrid model combining convolutional and recurrent layers achieved 95% accuracy, with precision of 0.88 and recall of 0.85, when classifying AF from PPG signals.6PubMed Central. A Deep Learning Approach for Atrial Fibrillation Classification Using Multi-Feature Time Series Data from ECG and PPG The consistent theme across these studies is that the raw waveform contains substantially more information than simple heart rate statistics, and modern algorithms know how to extract it.
Premature Beats and Ectopic Rhythms
AF is not the only arrhythmia a PPG can catch. Premature ventricular contractions and premature atrial contractions (PACs) each leave their own signatures. PVCs tend to produce a weak pulse followed by a compensatory pause and a strong pulse, while PACs create subtler timing irregularities that can be harder to spot visually but are still detectable algorithmically.
A hierarchical rule-based approach to detecting PVCs from PPG was recently shown to perform on par with more complex machine learning methods. By using percentage-based thresholds for beat-to-beat timing changes and pulse amplitude shifts rather than absolute values, the algorithm adapted to individual patients’ baseline waveforms. It achieved F1 scores of about 97% for one PVC pattern and 92% for a second, with corresponding Matthews correlation coefficients of 0.97 and 0.93.7Elsevier / Biomedical Signal Processing and Control. A hierarchical rule-based algorithm for detecting premature ventricular contractions (PVCs) in the photoplethysmogram That level of accuracy suggests even relatively simple algorithms can reliably flag PVCs when given clean PPG data.
When It Is Not Your Heart
This is where the practical difficulty lives. A jittery waveform on a pulse oximeter is far more commonly caused by finger movement, loose sensor contact, or shivering than by an actual arrhythmia. Motion artifacts are the single biggest source of false alarms, and they are deceptively similar to real rhythm disturbances. Movement shifts the sensor’s contact pressure on the skin, deforms the tissue underneath, and creates pulse-to-pulse interval variability that mimics the irregular spacing of atrial fibrillation.8npj Digital Medicine. Photoplethysmography based atrial fibrillation detection: a review
Researchers have found a clever workaround in wearable devices that also contain an accelerometer, the same motion sensor your phone uses to count steps. By feeding the accelerometer data into the classification model alongside the PPG, the algorithm can tell whether timing irregularities correlate with wrist or body movement. One study found that adding accelerometer input improved sensitivity for AF and premature beat detection by an average of about 4% and 6%, respectively. More importantly, the accelerometer helped the model correctly override waveform disturbances that coincided with physical motion, preventing them from being labeled as arrhythmias.9PubMed Central. Multiclass Arrhythmia Classification using Multimodal Smartwatch Photoplethysmography Signals Collected in Real-life Settings A similar approach used a denoising step that first identified heavily motion-corrupted segments by analyzing accelerometer data and then filtered or excluded them before running the AF classifier.10Expert Systems with Applications. Atrial fibrillation detection on reconstructed photoplethysmography signals collected from a smartwatch using a denoising autoencoder
Beyond motion, low perfusion can also flatten or distort the waveform. Cold fingers, low blood pressure, vasoconstricting drugs, peripheral vascular disease, and even pain or stress all reduce the perfusion index, making the signal noisy and harder to interpret.11PubMed Central. Peripheral perfusion index of pulse oximetry in adult patients: a narrative review If you are looking at your own oximeter at home and the waveform looks abnormal, warming your hand, staying still for 30 seconds, and repositioning the sensor will eliminate most non-cardiac causes.
How Consumer Smartwatches Perform
Modern smartwatches use the same PPG technology as medical pulse oximeters, just miniaturized and worn on the wrist. The obvious question is whether they are accurate enough to trust for arrhythmia screening. The research is increasingly reassuring. A systematic review and meta-analysis covering 26 studies and more than 17,000 patients found that smartwatches overall achieved 95% sensitivity and 97% specificity for atrial fibrillation detection, with a pooled area under the curve of 0.97. Among specific devices, the Apple Watch showed 94% sensitivity and 97% specificity, Samsung devices reached 97% sensitivity and 96% specificity, and the Withings ScanWatch came in at 89% sensitivity and 95% specificity.12PubMed Central. Accuracy of Smartwatches in the Detection of Atrial Fibrillation: A Systematic Review and Diagnostic Meta-Analysis
A separate meta-analysis compared PPG-based smartwatches directly against ECG-based smartwatches and found that the PPG approach was actually more sensitive, with pooled sensitivity of about 97% versus 83% for ECG-based devices.13PubMed Central. Comparison of diagnostic accuracy of electrocardiogram-based versus photoplethysmography-based smartwatches for atrial fibrillation detection: A Systematic Review and Meta-Analysis That finding may seem counterintuitive, since ECG is considered the gold standard. The likely explanation is that PPG-based passive monitoring can sample heart rhythm continuously throughout the day, catching intermittent AF episodes that a single on-demand ECG reading might miss. Another meta-analysis comparing PPG smartwatches to ECG chest patches confirmed the PPG devices’ slightly higher sensitivity, while chest patches had a marginal edge in specificity.14PubMed Central. Diagnostic accuracy of ECG smart chest patches versus PPG smartwatches for atrial fibrillation detection: a systematic review and meta-analysis
A caveat worth knowing: PPG-based heart rate variability measurements from the wrist agree best with chest-strap ECG readings when you are lying down and still. Agreement remains good while seated but degrades during activity. During sleep, correlations between wrist PPG and chest ECG for key variability metrics exceeded 0.9, while during high-motion periods accuracy dropped substantially.15Smart Wearable Technology. Assessing Pulse Rate Variability from a Wrist-Worn PPG Device Against ECG-Derived Heart Rate Variability in Ambulatory Settings In a controlled study comparing chest-strap and wrist sensors, the two showed excellent agreement for variability measures while supine, with intraclass correlations above 0.95, but agreement was somewhat lower in the seated position.16PubMed Central. A Comparative Study Between ECG- and PPG-Based Heart Rate Sensors for Heart Rate Variability Measurements: Influence of Body Position, Duration, Sex, and Age The practical takeaway: if your smartwatch flags an irregular rhythm while you are sitting quietly or sleeping, that alert deserves attention. If it flags something while you are running or waving your arms, take it with more skepticism.
Advanced Waveform Features and Derivative Analysis
Most consumers see only the basic waveform on a pulse oximeter screen, but researchers extract far more from the same signal. By computing the first and second mathematical derivatives of the PPG, scientists can identify subtle inflection points that mark events like the peak rate of blood flow into the finger, the moment arterial pressure starts to fall, and reflected pressure waves bouncing back from the periphery. These fiducial points are essential for calculating parameters used to diagnose cardiovascular conditions beyond arrhythmias, including arterial stiffness and vascular aging.17PubMed Central. Analysis on Four Derivative Waveforms of Photoplethysmogram (PPG) for Fiducial Point Detection You will not encounter derivative analysis on a bedside monitor, but it is the engine behind many of the algorithms running inside wearable devices and hospital monitoring systems.
Waveform Clues in the ICU
In critical care settings, the oximeter waveform takes on a different role: predicting whether a patient’s circulation will respond to intravenous fluids. The pleth variability index (PVI) measures how much the waveform’s amplitude swings with each breath. In a mechanically ventilated patient, large swings suggest the heart is underfilled and likely to pump more blood if given fluids. In one operating-room study, a PVI above 14% before fluid administration predicted a positive response with 81% sensitivity and 100% specificity.18British Journal of Anaesthesia. Pleth variability index to monitor the respiratory variations in the pulse oximeter plethysmographic waveform amplitude and predict fluid responsiveness in the operating theatre
In ICU patients receiving vasopressor medications like norepinephrine, the relationship between PVI and actual hemodynamic changes was weaker but still present. A PVI threshold of 11% discriminated higher from lower pulse pressure variation with about 70% sensitivity and 71% specificity, and the overall area under the curve was 0.80.19PubMed Central. Impact of norepinephrine on the relationship between pleth variability index and pulse pressure variations in ICU adult patients The lesson here is that vasopressors and other factors that alter vascular tone can blunt the waveform’s ability to reflect what the heart is actually doing. Clinicians interpreting the PVI in critically ill patients need to account for the medications in the picture.
How Skin Pigmentation Affects the Signal
Pulse oximeters work by shining light through tissue and measuring how much is absorbed by pulsating blood. Melanin in the skin also absorbs light, and in people with darker skin the increased absorption and altered light scattering can change the signal characteristics. Melanin increases the baseline absorption of light by the epidermis, and the scattering pattern of photons shifts toward a more forward direction in darker skin.20PubMed Central. A review of the effect of skin pigmentation on pulse oximeter accuracy The result is that standard oximeter calibrations, developed across populations that may not fully represent darker skin tones, can perform differently in these patients.
For oxygen saturation readings, this has been widely documented: pulse oximeters tend to overestimate oxygen levels in people with darker skin, sometimes masking dangerously low readings. For waveform-based arrhythmia detection, the impact is less studied but still relevant. A weaker or noisier signal at the sensor means algorithms have less clean data to work with, potentially increasing the chance of both missed arrhythmias and false alarms. If you have darker skin and are using a pulse oximeter or smartwatch for rhythm monitoring, be aware that signal quality may be lower and that persistent waveform irregularities are worth investigating with a clinical ECG rather than dismissed as noise.
Screening Newborns With the Waveform
One of the more creative uses of oximeter waveform analysis is in neonatal screening for coarctation of the aorta, a structural narrowing of the body’s main artery. By comparing the shape of the PPG waveform between a newborn’s hand and foot, researchers identified that babies with coarctation had a significantly shallower rate of waveform decline in the lower extremities. When comparing upper and lower limb waveforms, the difference in the maximum rate of fall was statistically significant, and the disparity resolved after surgical repair. The screening algorithm achieved an overall detection accuracy of 72%, with 61% sensitivity and 94% specificity.21PubMed Central. Using pulse oximetry waveforms to detect coarctation of the aorta The specificity is high enough to avoid flooding neonatal units with false alarms, and the approach requires no equipment beyond the pulse oximeters already clipped to every baby in the nursery. The sensitivity is moderate, meaning it would miss some cases, but as a layer added to existing screening it could catch defects that oxygen saturation alone overlooks.
This neonatal application highlights something broader about the field: oximeter waveform analysis is not limited to arrhythmias. The same pulsatile signal that reveals irregular heartbeats can also expose structural heart problems, vascular stiffness, fluid status, and autonomic dysfunction. The hardware has been around for decades. What is changing is the software and the recognition that the waveform was always carrying more information than anyone bothered to read.