Hyperspectral Wound Healing Smartphone

By /

Hyperspectral Wound Healing Smartphone

ISEF Category: Biomedical Engineering

Ready to Turn This Idea Into a Real Project?

This guide was put together with the help of AI research tools to give you a solid starting point. But a competitive science fair project lives in the details: refining your research question, fine-tuning your variables, analyzing your data, and presenting your findings like a seasoned scientist.

For next steps tailored to your interests, skill level, and timeline, work one-on-one with a MehtA+ mentor. Learn more about MehtA+ Science & Engineering Research Mentorship →

Subcategory: Biomedical Sensors and Imaging  ·  Difficulty: Advanced  ·  Setup: Home Setup  ·  Time: 1 to 2 Months

The Hook

Wound clinics judge healing by eye, which means experienced nurses do better than rookies and nobody can tell the difference for an itchy patch at home. Hyperspectral imaging splits each pixel into dozens of wavelengths and reveals chemical content directly. A 5-dollar grating on your phone is enough to demonstrate the workflow on public wound datasets.

What Is It?

Hyperspectral imaging records many narrow wavelength bands per pixel. The result is a 3D data cube. Bands related to oxyhemoglobin, deoxyhemoglobin, and water reveal tissue health.

A diffraction grating split-line attachment turns a single image into a smeared rainbow. With careful calibration, that smear becomes a pseudo-hyperspectral cube.

A CNN classifies wound regions into granulation (healthy), infection, or necrosis. Public NIH wound image datasets and published examples train and validate the model.

Why This Is a Good Topic

Hyperspectral imaging is a hot research area with very few high school projects. You will learn spectroscopy, calibration, and tissue classification with CNNs.

Research Questions

  • How does wavelength resolution change classifier accuracy?
  • What is the effect of ambient light on the spectral reconstruction?
  • Does hyperspectral input beat RGB-only baseline?
  • To what extent does skin tone affect classification bias?
  • Which wavelengths carry the most class-separating signal?
  • How does camera exposure shift apparent reflectance?
  • What is the effect of training-set size on cross-subject generalization?

Basic Materials

  • Diffraction grating film or piece of CD.
  • 3D-printed grating mount for the phone.
  • Smartphone with manual-exposure app.
  • Standardized lighting LED.
  • Public NIH wound image datasets with permissions.
  • Color-calibration card.

Advanced Materials

  • Lab hyperspectral camera.
  • Reference wavelength lamp.
  • Clinical mentor and wound-care dataset access.
  • Cloud GPU.

Software & Tools

  • PyTorch: Trains the wound classifier.
  • OpenCV: Reconstructs the spectral cube from grating images.
  • scikit-learn: Reports classification metrics and confusion matrices.
  • Roboflow: Manages dataset labeling.

Experiment Steps

  1. Calibrate wavelengths every session with a known lamp.
  2. Lock the lighting setup and color-card placement.
  3. Build a subject-wise split.
  4. Compare hyperspectral vs. RGB-only models.
  5. Train and validate.
  6. Audit fairness across skin tones and wound types.

Common Pitfalls

  • Skipping wavelength calibration between sessions.
  • Ignoring ambient light.
  • Treating high overall accuracy as proof on imbalanced classes.
  • Training and testing on the same patients.
  • Reporting visual examples without metrics.

What Makes This Competitive

A competitive entry runs subject-wise splits, performs wavelength calibration on every session, compares hyperspectral input vs. RGB-only, and audits fairness across skin tones. Validating model output against ground-truth wound classification by clinicians is the gold-standard step.

Project Variations

  • Apply the same pipeline to pressure-ulcer staging.
  • Replace CNN with a transformer and compare.
  • Add a smartphone IR sensor for thermal context.

Learn More

  • PubMed: Search hyperspectral wound classification review.
  • NIH PubMed Central: Open-access wound imaging papers.
  • OpenCV documentation: Free imaging tutorials.
  • NIST Hyperspectral Imaging Lab: Reference standards.
  • MIT OpenCourseWare: Course 6.815 Computational Photography.
Shopping Cart