Pulse Oximeter Calibration
Reference: Chapter 4 of Pulse Oximetry 2nd Edition by John TB Moyle
1. Overview
1.1 Introduction
Initial models of pulse oximeters relied on a theoretical framework built upon the Beer–Lambert law. The equation for this law is:
In this equation, OD stands for optical density, for incident light, for transmitted light, for the extinction coefficient, for sample concentration, and for optical path length.
However, the Beer–Lambert law is not practical for clinical pulse oximetry for several reasons. This theory only holds when the medium is non-turbid, contains a single solute, and has a constant optical path length. Because none of these conditions exist in clinical settings, manufacturers have turned to empirical methods for calibration. In the past, calibration often involved in vivo comparisons using arterial blood samples analyzed by CO-oximeters. Nowadays, teams are exploring in vitro calibration techniques and alternative non-hem methods.
1.2 What Do Pulse Oximeters Measure?
The main contention is around what exactly pulse oximeters measure: functional or fractional oxygen saturation. Both terms have historical roots in how oxygen content in the blood was originally measured. The van Slyke method focused on measuring the actual volume of oxygen per unit volume of blood. Oxygen saturation was then defined as a percentage of its capacity for oxygen. If we go by this definition, unreactive forms of hemoglobin like carboxyhemoglobin and methemoglobin are not included. The equation for functional hemoglobin saturation is:
Current advanced spectrophotometers can measure all commonly occurring forms of hemoglobin. The fractional saturation then becomes:
Ideally, pulse oximeters should clarify if they measure functional or fractional saturation. Unfortunately, with current two-wavelength designs, they don't precisely measure either. They use the absorption rates at two wavelengths to estimate an oxygen saturation value, termed SpO2.
1.3 Blood Gas Analyzers
Blood gas analyzers use electrodes to measure the partial pressures of oxygen and carbon dioxide, along with hydrogen ion concentration. Typically, oxygen is gauged using a Clark polarographic electrode, and carbon dioxide by a Severinghaus electrode. These devices also display oxygen saturation (SO2), but this value is inferred from measured oxygen pressure (PO2) and adjusted for variables like carbon dioxide pressure (PCO2), pH, and temperature.
However, even with these adjustments, accurately determining the percentage of oxygen saturation remains a challenge due to varying algorithms and the need to predict the oxyhemoglobin dissociation curve for each blood sample. The curve's shape can change due to pH, temperature, and PCO2 levels.
2. The "Gold Standard" in Pulse Oximetry Calibration
The term "gold standard" in pulse oximetry calibration refers to the CO-oximeter. This technology was first introduced in 1966 by Instrumentation Laboratories Inc.
2.1 Overview
How the CO-oximeter Works
The CO-oximeter measures oxygen saturation levels using small samples of arterial blood treated with heparin. Unlike pulse oximeters that give real-time data, the CO-oximeter provides a reading only for the moment the blood sample is taken. The device employs a technique similar to pulse oximetry, but the wavelengths it uses differ from those used in most pulse oximeters. While pulse oximeters usually use red and near-infrared wavelengths, CO-oximeters typically operate in the visible spectrum.
To obtain a reading, a 100 μl sample of heparin-treated blood is placed into the CO-oximeter. This sample is broken down by ultrasound to create a hemoglobin solution. The solution is then moved to a cuvette by a pump. Light from a tungsten-halogen lamp passes through a monochromator, which narrows the light to a specific range of wavelengths between 500 nm and 700 nm. This light goes through the cuvette containing the treated blood and is eventually detected by a photodiode. A microprocessor then applies the Beer–Lambert law and additional calculations to arrive at the oxygen saturation levels.
What It Measures
The CO-oximeter provides multiple readings, including: - Oxyhemoglobin - Carboxyhemoglobin - Methemoglobin - "Free" hemoglobin - Sulph hemoglobin
From these, it also calculates: - Total hemoglobin - Fractional oxygen saturation - Functional oxygen saturation - Oxygen capacity - Oxygen content
Accuracy and Variations
The AVL 912 CO-oximeter, for instance, uses up to 17 wavelengths to account for other factors that might affect the reading, like fetal hemoglobin and bilirubin. However, some CO-oximeters, like the Radiometer OSM3 Hemoximeter, use fewer wavelengths and smaller blood samples, which may impact calibration accuracy.
Comparing CO-oximeters and Pulse Oximeters
While both devices measure oxygen saturation, there are some key differences: - CO-oximeters work with hemoglobin solutions, not whole blood. - The optical path length in CO-oximeters is fixed. - CO-oximeters use a wider range of wavelengths. - The Beer–Lambert law is more directly applicable in CO-oximeters.
2.2 In vivo calibration of pulse oximeters
Pulse oximeters made before 1993 were calibrated in a live environment. During design and development, these devices are calibrated multiple times against a CO-oximeter to determine the relationship between oxygen saturation and light absorption at two wavelengths. After this, only occasional two-point checks are needed for each new pair of light-emitting diode (LED) and photodetector. Usually, no further checks are needed unless there is a significant change in the LED or detector.
Understanding the calibration process is crucial for users and buyers. A healthy, non-smoking volunteer with specific hemoglobin levels is usually chosen. The subject is exposed to different oxygen levels while their blood samples are analyzed with a CO-oximeter. This helps in plotting a calibration curve.
However, this process raises ethical concerns, especially since it involves lowering the oxygen levels in a healthy person's blood, sometimes posing health risks. In the U.S., there was a case where a volunteer experienced a seizure 24 hours after the test. As a result, the calibration curve often needs to be extrapolated for readings below 85% oxygen saturation.
Another issue involves the number and placement of calibration points. More data points between oxygen saturation levels of 95% to 100% aren't as useful as well-spaced points ranging from 100% to 80%.
For the calibration data to be universally applicable across a specific model, the LEDs and photodetectors in all units must be identical. Manufacturers either select LEDs with closely-matched wavelengths or adjust the calibration according to the wavelengths.
Operating current and temperature can affect the LED's wavelength, but these changes are usually negligible in affecting the oximeter's accuracy. Variability also arises due to different finger and earlobe thicknesses and pigmentation. Most devices automatically adjust LED intensity to maintain a stable signal. However, failure to compensate for changes in LED output can result in minor errors in low oxygen saturation levels, although the impact is minimal at higher saturation levels.
3. Accuracy in Pulse Oximetry
3.1 How Accurate Are They?
Pulse oximeters are calibrated using real-world data from healthy adults. The process has been found to be good enough for everyday clinical use. However, the accuracy varies among different brands, particularly when oxygen saturation is below 85%. This is because it's not ethical to calibrate these devices below 80% saturation using human subjects. In general, expect a deviation of 2-3% for saturations between 70-100%.
Mannheimer and his team found that certain LEDs offer better accuracy at different saturation levels. Additionally, your blood's composition, like its haematocrit level, can also impact the device's accuracy. More on issues that could affect accuracy will be discussed in Chapter 10.
3.2 Calibrating Without Human Subjects
Using non-human methods for calibration would offer several advantages, including:
- Ability to test below 85% saturation
- Easier repetition of studies
- No health risks
- Easier to study abnormal blood conditions
- Allows for standardization
Despite these advantages, current non-human test devices have limitations. They can only test specific points, and can't compare different brands of pulse oximeters.
Fisher and his team have developed a dye-based test device that performs well across various types of pulse oximeters. This kind of advancement is crucial since many studies point to increasing inaccuracies at low oxygen saturation levels.
Reynolds and colleagues developed a sophisticated system for this purpose, which involves circulating heparin-treated blood around a closed loop. This system successfully mimics human physiology and has been useful in testing different pulse oximeters. The following image shows the setup:
3.3 Simple Calibration Tools
There are also simpler devices for quick calibration checks. One device requires small blood samples and involves squeezing a bulb to simulate a pulse. While simpler, these methods have proven effective in providing reliable results.
4. Impact of Haemoglobin Levels on Pulse Oximetry Accuracy
In theory, pulse oximetry should work well regardless of varying haemoglobin concentrations. However, in practice, it's not that straightforward. Experiments using a simplified finger model showed that as haemoglobin levels moved away from 10 g/dL, the accuracy began to suffer. Specifically, the SpO2 measurements were too high with polycythaemia and too low with anaemia. The inaccuracies got worse as oxygen saturation levels decreased.
In dog experiments, other cardiovascular factors were kept stable while the packed cell volume was reduced. It was found that pulse oximetry and continuous mixed venous oxygen saturation remained clinically reliable as long as the packed cell volume stayed above 15%. However, low haematocrit levels led to some scatter and an underestimation of saturation.
When aggregating results from 43 pulse oximeters from 12 different manufacturers, a pattern emerged: The negative bias in SpO2 measurements increased in a near-linear fashion as haemoglobin concentrations dropped from 14g/dL to 8-9g/dL, especially when mean SaO2 was around 53.6%.
Interestingly, we still don't know the lowest haemoglobin level at which pulse oximetry becomes completely unreliable. Some studies have shown reasonable performance even at haemoglobin levels as low as 2.3 g/dL.