A wandering baseline on an ECG is a slow, rolling drift of the entire tracing up or down from where it should sit, making the flat segments between heartbeats look like gentle waves instead of a steady horizontal line. It is one of the most common artifacts in electrocardiography, and while it does not come from the heart itself, it can obscure findings that do. The drift is technically a low-frequency disturbance, usually below about 0.5 to 0.7 Hz, and it arises primarily from breathing, body movement, and problems at the skin-electrode interface.1PubMed Central. Comparison of Baseline Wander Removal Techniques considering the Preservation of ST Changes in the Ischemic ECG: A Simulation Study The artifact is so pervasive across recording devices that removing it is considered an unavoidable step before any serious ECG analysis can begin.
What the Baseline Is and Why It Drifts
On a normal ECG, there is an imaginary flat reference line, the “isoelectric line,” that represents the heart at electrical rest between beats. Every wave the ECG records, the P wave, QRS complex, T wave, is measured as a deflection above or below that line. When everything is working well, those resting segments stay put. A wandering baseline means this reference line itself is slowly rising and falling, as though the entire strip were printed on a sheet of paper being gently tilted back and forth. The clinical trouble is obvious: if the floor keeps moving, you cannot tell whether a wave is genuinely elevated or depressed, or whether the floor just shifted underneath it.
The most common cause is ordinary breathing. When you inhale and exhale, several things change at once. The electrical resistance of your lungs shifts as they fill with air. Your heart physically moves inside the chest as the diaphragm drops and rises. And the orientation of the heart’s electrical axis changes slightly with each breath cycle. All of these factors alter the voltages the ECG electrodes pick up, producing a slow oscillation in the tracing that tracks your respiratory rate.1PubMed Central. Comparison of Baseline Wander Removal Techniques considering the Preservation of ST Changes in the Ischemic ECG: A Simulation Study Since most people breathe somewhere between 12 and 20 times per minute, the drift tends to cycle at roughly 0.2 to 0.33 Hz, well below the frequencies of the heart’s own electrical signals.
Other Causes Beyond Breathing
Respiration is the biggest contributor, but it is not the only one. Patient movement during the recording, even small shifts like adjusting an arm or tensing a leg, can produce sudden jumps or slower undulations in the baseline. These motion artifacts can be dramatic enough to mimic abnormal beats. One review of ECG artifacts noted that limb movement during testing creates sudden irregularities in the baseline that may resemble premature contractions or even mimic supraventricular and ventricular arrhythmias.2PubMed Central. Main artifacts in electrocardiography A clinician who does not recognize the artifact might misinterpret it as a rhythm disturbance.
The electrode-skin interface is another frequent culprit. Perspiration can change the electrical properties of the contact point. Poor electrode adhesion, dried-out gel, or skin that was not properly prepped before electrode placement all increase impedance between the electrode and the skin, which shows up as low-frequency drift. Electrode type matters too: research comparing textile-based electrodes with standard silver/silver chloride gel electrodes found that the textile versions produced significantly more noise in the very low frequency band where baseline wander lives (below about 0.67 Hz), suggesting that electrode design directly influences how much drift you see.3Journal of Electrocardiology. A comparison of conductive textile-based and silver/silver chloride gel electrodes in exercise electrocardiogram recordings
Exercise testing deserves a special mention. When someone is walking or running on a treadmill, breathing is heavier and faster, the torso is in motion, and sweat increases. All of these amplify baseline wander simultaneously, which is why exercise ECGs have historically been among the hardest recordings to interpret cleanly.
Why It Matters for Diagnosis
Baseline wander is not just a cosmetic annoyance. It directly threatens the reliability of one of the most clinically important parts of the ECG: the ST segment. The ST segment sits between the QRS complex and the T wave, and its position relative to the baseline is the primary marker clinicians use to diagnose a heart attack or ongoing cardiac ischemia. A segment that is elevated or depressed by even one or two millimeters can trigger an emergency response. When the baseline itself is drifting by that amount or more, real ST changes can be hidden, and normal ST segments can look abnormal.1PubMed Central. Comparison of Baseline Wander Removal Techniques considering the Preservation of ST Changes in the Ischemic ECG: A Simulation Study
The risk runs in both directions. A wandering baseline that lifts the trace in certain leads can make it look like there is ST elevation when the heart is fine. A downward drift can mask genuine elevation. In an emergency department, where minutes count and a “STEMI” diagnosis (ST-elevation myocardial infarction) can send a patient straight to the catheterization lab, that kind of ambiguity has real consequences. The artifact can mean either a missed heart attack or an unnecessary invasive procedure.
The Filter Problem
The obvious solution to baseline wander is to filter it out. Since the drift sits at very low frequencies, you can use a high-pass filter that blocks everything below a certain cutoff and lets the heart’s faster signals through. Most modern ECG machines do this automatically. But here is where the cure can introduce its own disease: if the filter’s cutoff frequency is set too high, it does not just remove the drift. It also chops into the low-frequency components of the ST segment itself, distorting the very measurement you are trying to protect.
A recent study quantified just how dangerous this can be. When a 1.0 Hz high-pass filter was applied, the odds of a false-positive ST elevation of 1.0 mm or more increased nearly tenfold in lead V1 and about fivefold in lead V2, compared with a properly set lower-frequency filter.4PubMed. Quantifying high-pass filter-induced ST-segment distortion That is a striking finding: the very tool designed to clean up the tracing was generating phantom heart-attack signatures. This is why international guidelines recommend that diagnostic ECG recordings use a high-pass filter no higher than 0.05 Hz, or at most 0.67 Hz for monitoring settings. The 1.0 Hz setting, sometimes used in ambulatory or exercise contexts to aggressively suppress drift, should not be relied upon for ST-segment interpretation.
The tension between effective drift removal and signal preservation is the central engineering challenge in ECG processing, and it has driven decades of increasingly sophisticated approaches.
How ECG Machines Remove Baseline Wander
Three broad families of correction techniques have been developed, each with trade-offs that determine when it is used.
The simplest is the high-pass filter already described. A linear-phase high-pass filter with a cutoff frequency set below the heart rate can strip out the very slow drift while leaving most of the cardiac signal intact.5Journal of Electrocardiology. Filters for the reduction of baseline wander and muscle artifact in the ECG The Butterworth filter, a standard design, is extremely fast, with one simulation study reporting median processing times of about 0.006 seconds per ECG segment.1PubMed Central. Comparison of Baseline Wander Removal Techniques considering the Preservation of ST Changes in the Ischemic ECG: A Simulation Study Speed matters in clinical settings where ECGs are processed continuously, but as noted above, simple filters risk distorting the ST segment if the cutoff is not carefully chosen.
The second approach is cubic spline interpolation, which has been used for roughly four decades. Instead of filtering frequencies, this method identifies specific reference points on each heartbeat where the signal should be at the baseline, then draws a smooth curve through those points. Whatever that curve shows is treated as the estimated drift, and it is subtracted from the original signal. Because the spline uses knowledge of the ECG waveform’s shape rather than just its frequency content, it can estimate the true baseline more accurately and avoid some of the distortions that filters introduce.6Journal of Electrocardiology. Problems and limitations of ECG baseline estimation and removal using a cubic spline technique during exercise ECG testing: Recommendations for proper implementation The catch is that picking the right reference points matters enormously. Get them wrong, perhaps because the algorithm mistakes a noisy segment for a true isoelectric point, and the correction itself becomes inaccurate.7PubMed. An Improved Cubic Spline Interpolation Method for Removing Electrocardiogram Baseline Drift
The third and increasingly dominant family uses wavelet transforms. These break the ECG signal into components at different scales, identify the very-low-frequency components that correspond to baseline wander, and remove them while leaving the rest of the signal essentially untouched. A head-to-head comparison of 14 different wavelet types found that the best-performing wavelets preserved the original ECG morphology with a correlation coefficient above 0.99 and introduced virtually zero distortion to the ST segment, with a median deviation of 0.00 millivolts.1PubMed Central. Comparison of Baseline Wander Removal Techniques considering the Preservation of ST Changes in the Ischemic ECG: A Simulation Study A separate evaluation of wavelet transforms across different types of simulated drift confirmed strong performance for smooth, sinusoidal wander but noted that sudden shifts, like spikes or step changes, were harder to remove cleanly, though the wavelets could at least pinpoint where the sudden jump occurred.8PubMed Central. Comparing different wavelet transforms on removing electrocardiogram baseline wanders and special trends
More recent research has explored adaptive signal-piloted filters that adjust their own parameters in real time as the incoming ECG signal changes. One such system achieved roughly double the computational efficiency of conventional approaches while maintaining comparable output quality, making it potentially useful for portable or resource-constrained devices.9PubMed Central. Baseline wander and power-line interference elimination of ECG signals using efficient signal-piloted filtering
Wandering Baseline Versus Other ECG Artifacts
Baseline wander is not the only thing that can make an ECG look wrong. It is worth knowing how it differs from other common artifacts, because the fixes are different and the diagnostic risks are different.
Power-line interference, sometimes called 50 or 60 Hz noise depending on where you are, produces a rapid, fine-grained fuzz superimposed on the tracing. It comes from nearby electrical equipment and power cables. It looks nothing like the slow, rolling drift of baseline wander; instead, the trace appears thickened or fuzzy without shifting up or down. A simple notch filter targeted at the power-line frequency can remove it without affecting the cardiac signal much at all.
Muscle artifact (also called EMG artifact) comes from the electrical activity of skeletal muscles, particularly when a patient is tense, shivering, or has tremor. It occupies much of the same frequency range as the heart’s own signals, which makes it far harder to separate with standard filtering.5Journal of Electrocardiology. Filters for the reduction of baseline wander and muscle artifact in the ECG Where baseline wander is slow and sweeping, muscle artifact is rapid and jagged. In practice, the two often appear together, especially during exercise testing, and each demands a different correction strategy.
Electrode pop or lead disconnect produces a dramatic, sudden spike or a flat line when a lead loses contact entirely. This is usually easy to spot, since it is abrupt rather than gradual. But the moment right after reconnection can produce a transient baseline shift that overlaps with ordinary wander, so it sometimes takes a trained eye to sort out what happened.
Practical Steps to Minimize Drift Before Recording
Much of the fight against baseline wander happens before the recording even starts. Good electrode preparation is the single most effective preventive measure. That means cleaning the skin with an alcohol wipe, lightly abrading if needed to reduce impedance, applying fresh electrodes with intact gel, and pressing them firmly into place. Dried-out or reused electrodes are a common source of drift in busy clinical settings.
Patient positioning and instruction help too. Having the person lie still and breathe normally, or briefly hold their breath for a short strip, can dramatically reduce respiratory artifact. During exercise testing, where you cannot ask someone to hold still, electrode placement on the torso rather than the limbs helps reduce motion artifact, and ensuring the cables are secured so they do not swing or pull on the electrodes matters more than people tend to realize.
On the equipment side, making sure the ECG machine is properly grounded, that cables are not running alongside power cords, and that the device’s filters are set appropriately for the clinical context (diagnostic mode versus monitoring mode) are standard best practices. Diagnostic mode uses lower filter cutoffs that preserve ST-segment fidelity at the cost of tolerating slightly more visible drift; monitoring mode uses higher cutoffs that produce a cleaner-looking trace but, as discussed, risk distorting the very findings that matter most.
Artificial Intelligence and ECG Quality
The latest frontier in managing baseline wander and other artifacts involves machine learning. AI systems are being developed both to remove noise from ECG recordings and to flag recordings that are too noisy to trust. One approach trains deep-learning models to act as quality gatekeepers, classifying ECGs as adequate or poor based on the presence and severity of artifacts like baseline wander and power-line interference.10Circulation. Abstract 4366206: ECG-IQ: Deep Learning-Enabled System that Defines Appropriateness of ECG Image Quality Enhances the Application of AI-ECG Diagnostics at the Point of Care The idea is that if an AI diagnostic tool is about to analyze an ECG, it should first verify that the recording is clean enough for the analysis to be meaningful.
On the denoising side, AI is being applied to problems that classical filters struggle with. Traditional filtering and wavelet methods have trouble when multiple types of noise overlap or when the noise pattern is unusual. Autoencoder-based neural networks, which learn to reconstruct a clean signal from a noisy input, have shown promise in suppressing diverse noise patterns while preserving the subtle morphological features that clinicians need to see.11arXiv. Enhancing AI-Based ECG Delineation with Deep Learning Denoising Techniques More broadly, AI systems are being used not just to clean up signals but to extract features invisible to the human eye, including extremely subtle variations in beat-to-beat timing and waveform shape that sit beneath the noise floor of a conventional reading.12PubMed Central. Current and Future Use of Artificial Intelligence in Electrocardiography
When a Wandering Baseline Should Prompt a Repeat Recording
Not every instance of baseline wander requires starting over. Mild drift that does not obscure the waveforms and can be corrected by the machine’s built-in filtering is usually acceptable, and most modern ECG reports are generated from digitally corrected signals anyway. But there are situations where the right call is to stop and re-do the recording.
If the drift is so severe that the P waves or ST segments are being pushed off-screen or into the adjacent channel’s territory, no amount of post-processing is going to reliably reconstruct what the heart actually did. If you can see that an electrode has come loose, is visibly dried out, or the patient was moving significantly during the recording, a repeat with fresh electrodes and the patient settled will always be more trustworthy than a digitally salvaged trace. In an acute chest-pain scenario, where ST-segment measurements directly determine treatment, even moderate baseline wander should lower your confidence in the recording, and a clean repeat strip takes only seconds.
The broader lesson is that baseline wander sits in an awkward middle ground: it is universally present to some degree, usually manageable with modern signal processing, but never completely harmless. Clinicians who read ECGs learn to glance at the overall quality of the trace before interpreting the waveforms, and a visibly wandering baseline is one of the first things that prompts either a mental adjustment or a request for a new recording. In an era when AI tools increasingly handle the initial interpretation of ECGs, the same quality-check step is being automated, ensuring that diagnostic algorithms do not confidently analyze a tracing that was too noisy to trust in the first place.