Smart Bandage Skin Impedance Classifier

ISEF Category: Embedded Systems

Subcategory: Sensors  ·  Difficulty: Advanced  ·  Setup: University Lab  ·  Time: Full Year

The Hook

A bandage can do more than cover skin. It can also act like a tiny sensor patch that tracks how tissue changes as it heals. That means your project can turn invisible electrical signals into a simple color cue. You get a real medical-tech problem, plus a clear engineering challenge.

What Is It?

This project studies how skin impedance changes during wound healing and whether a small sensor system can sort those changes into healing stages. Skin impedance is the skin’s resistance to a tiny electrical signal. Dry, intact skin and healing tissue do not behave the same way, so the signal can shift as the wound environment changes.

Think of it like checking how easily water flows through different sponges. A dry sponge, a wet sponge, and a damaged sponge all let water move differently. Your bandage tries to read that difference with two flexible electrodes, an instrumentation amplifier such as the INA826, and a microcontroller that turns the measured signal into a simple classification and LED feedback.

The engineering challenge is not just measuring something. You need a sensor front end that can pick up small changes, a clean way to process noisy data, and a classification rule that works on the microcontroller. That makes this a strong embedded systems project, because you are building both the sensing hardware and the decision logic.

Why This Is a Good Topic

This is a strong science fair topic because you can test a real signal, not just build a gadget. You can vary electrode design, placement, signal frequency, or classification method and measure how each choice affects accuracy and stability. The topic connects to wound care, wearable sensors, and low-cost health monitoring. You can learn sensor conditioning, data collection, feature extraction, and simple embedded classification in one project.

Research Questions

• How does electrode spacing affect the stability of skin impedance readings?
• What is the effect of flexible electrode material on signal noise and repeatability?
• Does skin impedance change enough across simulated healing states to support classification?
• To what extent does signal frequency change the separation between impedance measurements from different skin conditions?
• Which feature set gives the best healing-stage classification on a microcontroller?
• How does motion or bandage pressure affect classification accuracy?

Basic Materials

• Flexible conductive electrodes or conductive fabric strips.
• INA826 instrumentation amplifier or similar low-noise instrumentation amplifier.
• Microcontroller board with analog input, such as an Arduino, ESP32, or similar MCU.
• Breadboard or solderable prototyping board.
• Jumper wires and clips.
• Resistors and capacitors for amplifier gain and filtering.
• Stable power supply or USB power source.
• Multimeter.
• Disposable gloves.
• Notebook or lab record sheet.
• Nonconductive tape or medical-style adhesive backing for mounting.
• Reference materials for safe skin-contact testing and ethics review.

Advanced Materials

• Flexible printed circuit or custom sensor substrate.
• Precision resistor set for gain and calibration checks.
• Low-noise signal generator or function generator.
• Oscilloscope or logic analyzer with analog support.
• Data acquisition system for validation runs.
• Bioimpedance test phantom or saline-based tissue phantom for repeatable testing.
• Clinical-grade reference electrode materials for comparison.
• Shielded cables and grounded enclosure for noise control.
• PCB prototyping tools or custom board access.
• Temperature and humidity sensor for environmental logging.
• Higher-resolution ADC module if the MCU input is too noisy.
• Statistical analysis software for classification testing.

Software & Tools

• Arduino IDE: Programs the microcontroller and logs impedance features from the sensor front end.
• Python: Cleans data, plots impedance trends, and tests classification rules.
• Jupyter Notebook: Helps you explore signals, compare features, and document your analysis.
• ImageJ: Measures color changes from LED feedback photos if you want to quantify display output.
• Excel: Organizes trials, calculates summary statistics, and checks repeatability.

Experiment Steps

1. Define the exact sensing goal, such as separating simulated healing states or comparing skin conditions, before you build the circuit.
2. Choose one sensor geometry and one amplifier gain range, then plan how you will keep those settings fixed while testing.
3. Design a calibration plan that converts raw ADC counts or voltage changes into a comparable impedance feature.
4. Set up control samples or phantoms so you can compare the sensor against repeatable reference conditions instead of only live skin.
5. Decide which features the MCU will classify, then pick a simple decision rule or model that fits your memory and speed limits.
6. Plan a validation test that checks repeatability, motion sensitivity, and false-trigger behavior before you claim the system works.

Common Pitfalls

• Using poor electrode contact, which makes impedance changes look like healing changes.
• Letting ambient electrical noise enter the amplifier, which can swamp the tiny biosignal.
• Skipping a calibration reference, which makes raw readings hard to compare across sessions.
• Testing only one skin condition, which can make the classifier look accurate without being meaningful.
• Ignoring motion or bandage pressure, which can cause false stage changes on the MCU.

What Makes This Competitive

A stronger version of this project goes beyond a demo that lights an LED. You need a clear sensing model, a repeatable test setup, and a validation plan that shows your system can tell signal change from noise. Competitive work also compares electrode designs, amplifier settings, or classification methods, then backs the choice with data. If you can test a novel wear-time or motion-noise angle, your project becomes much more interesting.

Project Variations

• Test the same sensing idea with simulated wound phantoms instead of skin, then compare how well each phantom mimics healing-stage impedance.
• Replace the LED output with wireless transmission, then study whether classification stays stable after signal packet loss or delay.
• Compare two feature pipelines, such as raw thresholding versus a simple ML model, to see which one works better on-device.

Learn More

• PubMed: Search review articles on bioimpedance, wound monitoring, and wearable sensors to learn the medical background.
• NIH PubMed Central: Read free full-text papers on skin impedance, instrumentation amplifiers, and biosignal classification.
• IEEE Xplore: Search abstracts and open papers on wearable bioimpedance sensors and embedded classification methods.
• NASA NTRS: Look for free engineering reports on low-noise sensor design and signal conditioning concepts.
• MIT OpenCourseWare: Find circuit analysis and signals courses that help with amplifier design and filtering.