You clip it on your finger, a red light glows, and a number appears on the screen. That number — your SpO2 reading — tells clinicians whether your blood is carrying enough oxygen. Pulse oximeters are everywhere: hospitals, ambulances, doctor's offices, and since COVID-19 prompted a run on home medical devices, millions of living rooms. They're cheap, noninvasive, and provide continuous readings without drawing blood.
But the physics behind that little glowing clip is more interesting than most people realize, and the limitations are more serious than the medical device industry acknowledged for decades.
"Black patients had nearly three times the odds of occult hypoxemia that was not detected by pulse oximetry, compared with white patients." — Sjoding et al., New England Journal of Medicine (2020)
Two wavelengths, one ratio
The core principle is surprisingly elegant. Pulse oximetry exploits a physical property of hemoglobin: oxygenated hemoglobin (HbO2) and deoxygenated hemoglobin (Hb) absorb light differently at different wavelengths.
At 660 nanometers (red light), deoxygenated hemoglobin absorbs significantly more light than oxygenated hemoglobin. At 940 nanometers (infrared), the relationship flips — oxygenated hemoglobin absorbs more. By measuring the ratio of light absorption at these two wavelengths, the device can calculate what percentage of your hemoglobin is carrying oxygen.
The math works out to a ratio called R:
R = (AC_red / DC_red) / (AC_IR / DC_IR)
Where AC represents the pulsatile component (the signal that changes with each heartbeat as arterial blood volume fluctuates) and DC represents the non-pulsatile component (absorption from venous blood, tissue, bone, and skin pigment). This R value maps to an SpO2 percentage through empirically derived calibration curves — curves that were built by comparing pulse oximeter readings against arterial blood gas samples in human volunteers.
That calibration step matters more than it sounds. Those volunteers, historically, were predominantly light-skinned. We'll come back to that.
Isolating the arterial signal
One of the clever aspects of pulse oximetry is how it separates arterial blood from everything else. Your finger contains arteries, veins, capillaries, muscle, bone, and skin — all of which absorb light. The device only cares about arterial blood, because that's what carries fresh oxygen from the lungs.
The trick is that arterial blood is the only component whose volume changes with each heartbeat. When the left ventricle contracts, arterial blood volume in your finger increases slightly, absorbing a bit more light. Between beats, it decreases. This pulsatile signal — the AC component — rides on top of a much larger constant signal from everything else. By isolating just the fluctuating part, the oximeter effectively filters out all the non-arterial absorption.
This is photoplethysmography (PPG) in its most basic form: using light to measure volume changes in tissue caused by the cardiac cycle. Every pulse oximeter is fundamentally a PPG device.
What can go wrong
Pulse oximeters are reliable enough that clinicians depend on them for moment-to-moment patient management. But "reliable enough" has limits, and those limits have real clinical consequences.
| Factor | How It Affects Accuracy | Clinical Impact |
|---|---|---|
| Low perfusion (cold fingers, hypotension) | Weak pulsatile signal, poor signal-to-noise | Unreliable readings or signal dropout |
| Motion artifact | Patient movement corrupts the AC signal | False alarms, inaccurate readings |
| Ambient light interference | External light sources add noise to the detector | Erroneous SpO2 values |
| Nail polish / acrylic nails | Absorbs or blocks light transmission | Artificially low or erratic readings |
| Darker skin tones (Fitzpatrick V-VI) | Melanin absorbs light across measurement wavelengths | Overestimation of SpO2, missed hypoxemia |
| Carboxyhemoglobin (CO poisoning) | CO-bound hemoglobin absorbs like HbO2 at 660nm | Falsely normal SpO2 despite dangerous CO levels |
| Methemoglobin | Absorbs equally at both wavelengths | SpO2 reads ~85% regardless of true saturation |
| Severe anemia | Reduced hemoglobin concentration | Potentially inaccurate at very low Hb levels |
| SpO2 below 70% | Calibration data is sparse below this range | Accuracy degrades significantly |
The skin tone problem
The Sjoding et al. study in the New England Journal of Medicine (2020) put numbers to something researchers had documented for years: pulse oximeters systematically overestimate oxygen saturation in patients with darker skin. Black patients in their analysis were three times more likely to have occult hypoxemia — oxygen levels low enough to warrant intervention — that the pulse oximeter reading failed to detect.
In January 2026, an FDA-funded study published in the BMJ tested pulse oximeters across participants with varying skin pigmentation. The results were, as STAT News characterized them, "surprising and confusing" — some devices met accuracy standards across skin tones while others didn't, and the pattern wasn't consistent.
The underlying physics is straightforward. Melanin absorbs light broadly across the visible and near-infrared spectrum. This absorption adds to the DC component of the signal, altering the R ratio that the device uses to calculate SpO2. Since the calibration curves were built using predominantly lighter-skinned subjects, the mathematical relationship between R and actual oxygen saturation breaks down for people with more melanin.
The FDA issued updated guidance in 2025 requiring manufacturers to test pulse oximeters across a broader range of skin pigmentations. Whether this translates to meaningfully better devices remains to be seen.
From contact to contactless
Traditional pulse oximetry requires physical contact — a sensor clipped to a finger, toe, or earlobe. Remote photoplethysmography (rPPG) applies the same fundamental principle without touching the patient at all.
Instead of shining an LED through tissue and measuring transmitted light, rPPG uses an ambient or controlled light source and a camera to detect the tiny color changes in skin that occur with each heartbeat. The blood volume fluctuations that a pulse oximeter measures through your fingertip also happen in your face, and they're visible to a camera — barely, but measurably.
For heart rate, rPPG works reasonably well. Multiple studies have demonstrated mean absolute errors under 2 bpm compared to contact-based reference devices. SpO2 estimation via camera is harder. The signal is weaker, the wavelengths captured by standard RGB cameras don't perfectly match the red and infrared wavelengths optimized for pulse oximetry, and the noise floor is higher when you're measuring reflected light from a face across a room rather than transmitted light through a fingertip.
Van Gastel, Stuijk, and de Haan at TU Eindhoven have published some of the most significant work on camera-based SpO2, demonstrating feasibility under controlled conditions. But field accuracy — especially across skin tones and lighting conditions — remains an active area of research.
What pulse oximetry can and cannot tell you
A pulse oximeter gives you one number: the percentage of hemoglobin molecules carrying oxygen. It does not tell you total oxygen content in the blood (which depends on hemoglobin concentration), it cannot distinguish carboxyhemoglobin or methemoglobin from normal hemoglobin, and it provides no information about ventilation or CO2 levels.
In practice, SpO2 is a screening tool. Values above 95% are generally normal. Values between 90-95% warrant attention. Values below 90% typically require intervention. But these thresholds were established with the assumption that the measurement is accurate — an assumption that doesn't hold equally across all patients.
The technology works well for what it does. The problem is that what it does has blind spots that took decades to formally acknowledge, and those blind spots fall disproportionately on certain patient populations. Understanding how pulse oximetry works — really works, down to the physics and the calibration curves — is the first step toward building something better.
Frequently Asked Questions
What does a pulse oximeter actually measure?
A pulse oximeter measures the percentage of hemoglobin molecules in your arterial blood that are carrying oxygen, known as SpO2. It does this by shining two wavelengths of light — red at 660nm and infrared at 940nm — through your tissue and comparing how much each wavelength is absorbed.
How accurate are pulse oximeters?
FDA-cleared pulse oximeters are generally accurate within 2-3% for SpO2 readings between 70-100%. However, accuracy degrades at lower saturation levels, during motion, with poor peripheral perfusion, and across darker skin tones. A 2020 NEJM study found Black patients were three times more likely to have dangerously low oxygen levels missed by pulse oximetry.
Can a phone camera measure blood oxygen?
Smartphone cameras can estimate SpO2 using the same photoplethysmography principles as pulse oximeters, either by placing a finger over the camera lens or through remote photoplethysmography (rPPG) of the face. Heart rate measurement via phone cameras is relatively mature, but SpO2 estimation remains less reliable and is not yet suitable for clinical decisions.
Why do pulse oximeters work less well on darker skin?
Melanin in the skin absorbs light across the same wavelengths pulse oximeters use for measurement. Higher melanin content can alter the ratio of red to infrared light absorption, introducing bias into the SpO2 calculation. The calibration curves in most pulse oximeters were developed primarily using lighter-skinned subjects, compounding the problem.