Thermal Imaging Raynaud’s Rewarming Study

ISEF Category: Biomedical Engineering

Subcategory: Biomedical Sensors and Imaging  ·  Difficulty: Advanced  ·  Setup: School Lab  ·  Time: Full Year

The Hook

Your fingers can turn into tiny heat maps. That matters if you want to track how fast blood flow comes back after a cold trigger. A thermal camera can turn that recovery into data. You can then test whether the rewarming pattern predicts Raynaud's severity.

What Is It?

Raynaud's phenomenon is a condition where blood vessels in the fingers overreact to cold or stress. The fingers lose heat fast, then warm back up slowly. A thermal camera measures surface temperature, so you can watch that recovery happen frame by frame.

Think of it like watching ice melt, but for heat. Instead of guessing who rewarmed faster, you record the temperature of each finger over time. That gives you a curve, which is a graph of change over time. The shape of that curve can reveal how the body responds after cold exposure.

A recurrent network is a type of machine learning model that reads data in sequence. That makes it a good fit for rewarming curves, because the order of the temperatures matters. Your project links a body signal, thermal imaging, and prediction in one study.

Why This Is a Good Topic

This is a strong science fair topic because you can measure a real biological response, turn it into numbers, and test whether the pattern predicts a health outcome. It connects to circulation, medical imaging, and machine learning, so the project has both engineering and health value. You can start with a simple thermal camera study, then add deeper analysis if your data set grows.

Research Questions

• How does finger rewarming rate differ between people with and without a history of Raynaud's symptoms?
• What is the effect of finger position, such as index finger versus ring finger, on rewarming kinetics after cold exposure?
• Does baseline skin temperature predict the speed of recovery after a standardized cold-water challenge?
• To what extent do temperature curve features, such as slope and time to halfway recovery, classify Raynaud's severity?
• Which thermal features best separate mild, moderate, and severe Raynaud's patterns?
• How does a recurrent network compare with a simple regression model for predicting Raynaud's severity from rewarming curves?

Basic Materials

• MLX90640 thermal camera module or similar infrared array camera.
• Microcontroller or single-board computer that can read the thermal sensor.
• Stable camera mount or tripod for fixed hand positioning.
• Insulated container for a standardized cold-water challenge.
• Digital thermometer for water temperature checks.
• Timer or stopwatch.
• Computer for data logging and analysis.
• Consistent room with limited drafts and stable lighting.
• Disposable gloves for handling wet materials between participants.
• Consent and screening forms approved by your school or local review process.

Advanced Materials

• Higher-resolution thermal camera for comparison imaging.
• Reference blackbody source or calibrated temperature targets.
• Pulse oximeter for optional circulation comparison.
• Skin temperature probe for cross-checking thermal readings.
• Computer with GPU access for sequence modeling.
• Python environment for model training and validation.
• ImageJ for image inspection and region-of-interest checks.
• Statistical software or notebooks for mixed-effects analysis.
• Environmental monitor for room temperature and humidity.
• Custom hand-positioning jig for repeatable framing.

Software & Tools

• Python: Processes thermal frames, extracts finger temperature curves, and fits predictive models.
• ImageJ: Helps inspect thermal images and define consistent regions of interest on each finger.
• Jupyter Notebook: Organizes analysis, plots rewarming curves, and documents model tests.
• scikit-learn: Builds baseline classifiers and regressors for comparison with sequence models.
• PyTorch: Trains the recurrent network on temperature sequences.

Experiment Steps

1. Define the exact outcome you want to predict, such as symptom group, severity class, or rewarming speed.
2. Choose one thermal feature set first, then decide how you will convert images into one curve per finger.
3. Plan a repeatable cold challenge and a fixed camera setup so each trial starts from the same baseline.
4. Build your control plan, including healthy comparison participants, repeated trials, and environmental checks.
5. Decide how you will validate the model, including a train-test split, cross-validation, and a baseline method.
6. Set your analysis rules before collecting data, so you know which metrics will count as success.

Common Pitfalls

• Letting room temperature drift between sessions, which changes finger rewarming speed and hides the real pattern.
• Moving the camera or hand position between trials, which shifts the region of interest and corrupts the temperature curve.
• Using different cold challenge setups for different participants, which makes the exposures impossible to compare.
• Treating the hottest pixel on each finger as the full signal, which makes noise and reflections drive the results.
• Training the recurrent network on too few participants, which makes the model memorize people instead of learning Raynaud's patterns.

What Makes This Competitive

A class-level version of this project might stop at simple before-and-after temperature graphs. A stronger entry compares multiple feature sets, tests more than one model, and checks whether the model generalizes to new people. You can push it further by separating signal from noise with good controls, then asking which thermal features really carry clinical value. That turns the project from a demo into a real measurement study.

Project Variations

• Compare rewarming patterns in different fingers, such as thumb, index, and ring finger.
• Test whether fingertip color changes from a normal RGB camera add prediction power when combined with thermal data.
• Analyze whether simple curve metrics, such as peak drop and recovery slope, perform as well as a recurrent network.

Learn More

• PubMed: Search for review articles on Raynaud's phenomenon, thermal imaging, and cold challenge testing.
• NIH: Look for patient-friendly and clinical background on Raynaud's phenomenon and related circulation topics.
• NCBI Bookshelf: Read accessible textbook chapters on peripheral circulation, vasoconstriction, and biomedical imaging basics.
• NASA Earthdata and thermal remote-sensing resources: Study how infrared imaging measures temperature patterns, which helps with thermal camera thinking.
• MIT OpenCourseWare: Find free course materials on machine learning and signal processing to learn sequence modeling and validation.