Arduino Pulse Oximeter Bias Study

ISEF Category: Physics and Astronomy

Subcategory: Biological Physics  ·  Difficulty: Advanced  ·  Setup: School Lab  ·  Time: Full Year

The Hook

A pulse oximeter can look simple and still misread the same body in different conditions. That matters when doctors use it to make fast decisions. You can build a basic version, then test how light, motion, and skin-tone proxies change its signal. This turns a tiny sensor into a real fairness study.

What Is It?

A pulse oximeter estimates blood oxygen saturation, called SpO₂, by shining light through or onto skin and reading how much light comes back. The device watches how the signal changes with each heartbeat. Think of it like listening for one drum in a noisy room. The heartbeat is the drum, and ambient light, motion, and skin pigment are the background noise.

Your project asks a physics question, not just a health question. How do the sensor signal and the error floor change when the measurement gets harder? If you vary light conditions, movement, and skin-tone proxies, you can measure when the device starts to fail and how badly. Then you can test a signal-processing fix, like better filtering, motion rejection, or baseline correction, to see whether it lowers error across groups.

Why This Is a Good Topic

This is a strong science fair topic because you can measure clear inputs and outputs, and you can improve the design with data. The problem connects to real health equity, since pulse oximeter bias can affect care decisions. You can learn sensor physics, signal processing, calibration, and statistics without needing a full medical lab. You also get a real engineering angle, because you are not just measuring a problem, you are trying to reduce it.

Research Questions

• How does ambient light level change the noise floor of an Arduino photodiode pulse-oximeter clone?
• What is the effect of controlled motion on SpO₂ estimate error in a pulse-oximeter clone?
• Does adding a motion-rejection filter reduce SpO₂ error more than simple averaging?
• To what extent do skin-tone melanin proxies change the signal-to-noise ratio of reflected light measurements?
• Which wavelength pairing or optical geometry gives the smallest error across skin-tone proxies?
• How does baseline correction affect SpO₂ stability across different ambient light conditions?
• What is the effect of combining ambient-light subtraction and motion filtering on overall measurement bias?

Basic Materials

• Arduino board, such as an Uno or Nano.
• Photodiode sensor module or photodiode plus resistor circuit.
• Red and infrared LEDs.
• Breadboard and jumper wires.
• Finger clip or simple mounting holder.
• Opaque tape or dark enclosure material.
• Smartphone camera for documentation.
• Digital kitchen scale for repeatable setup checks.
• Colored filters, translucent tape, or printed skin-tone proxy cards.
• Laptop with Arduino IDE and spreadsheet software.

Advanced Materials

• Arduino board with analog input and stable power supply.
• Photodiode with transimpedance amplifier components.
• Red and infrared LEDs with current-limiting resistors.
• Optical filter set or interchangeable LED and detector mounts.
• Small accelerometer or IMU for motion tracking.
• Reference pulse oximeter for comparison data, if allowed by your lab.
• Neutral density filters for ambient-light testing.
• Dark box or optical bench fixtures.
• Human-subject consent materials, if your school and mentor approve.
• Data acquisition laptop with analysis software.

Software & Tools

• Arduino IDE: Programs the microcontroller and streams sensor readings to your laptop.
• Python: Cleans the data, applies filters, and runs error analysis.
• ImageJ: Measures color values if you use skin-tone proxy cards in photos.
• LibreOffice Calc: Organizes trial data and makes quick charts.
• R: Runs statistics, confidence intervals, and group comparisons.

Experiment Steps

1. Define the exact signal you will measure, such as raw photodiode output, pulse amplitude, or computed SpO₂ estimate.
2. Choose one main noise source to vary first, then plan separate tests for ambient light, motion, and skin-tone proxies.
3. Design a repeatable mounting setup so the sensor position stays fixed across trials.
4. Decide how you will convert raw sensor readings into a calibration metric, such as signal-to-noise ratio or percent error.
5. Plan control conditions that isolate each factor, so you can tell whether one variable caused the change.
6. Build a comparison plan for your signal-processing fix, then decide which metric will prove whether it helped.

Common Pitfalls

• Changing sensor distance from the skin between trials, which makes optical path length vary and hides the real effect.
• Mixing ambient light with LED signal because the detector is not shielded well enough.
• Testing motion with hand movement that is too inconsistent, which makes the noise hard to compare across runs.
• Using skin-tone proxies that change both color and reflectivity at once, which confounds the melanin proxy with texture effects.
• Comparing processed SpO₂ numbers without first checking whether the raw signal quality improved, which can hide a false fix.

What Makes This Competitive

A strong version of this project goes beyond a simple demo. You would quantify error with a real metric, compare more than one filtering method, and test whether the improvement holds across multiple noise sources. The best entries usually add a careful control design and a fair comparison across skin-tone proxies, not just one hand-picked example. If you also connect your results to published bias work, your project starts to look like research, not a class exercise.

Project Variations

• Test reflected-light versus transmitted-light sensor geometry to see which one is less sensitive to ambient light and skin-tone proxies.
• Compare simple moving-average filtering with adaptive filtering or IMU-based motion correction for reducing pulse-oximeter error.
• Swap skin-tone proxy materials, such as filters, printed cards, or fabric samples, to see how optical attenuation changes the measurement bias.

Learn More

• NIH PubMed: Search for review articles on pulse oximeter bias, skin pigmentation, and SpO₂ accuracy.
• FDA: Read safety and performance guidance for pulse oximeters in the device and medical device pages.
• Nature Medicine: Search for peer-reviewed studies on racial bias in pulse oximetry.
• MIT OpenCourseWare: Look for introductory electronics and signal processing materials that explain sensors, noise, and filtering.
• NASA NTRS: Search for signal-processing and photodetector references that help with noise reduction and optical measurements.