Smartphone Polarimetric Skin Hydration

ISEF Category: Biomedical Engineering

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

The Hook

Dry skin and connective-tissue disorders both change the way light bounces off your fingernails and palms. Polarized light pulls those changes out of the noise. Two cheap polarizer films and a phone camera become a polarimeter that maps birefringence. A regression CNN turns those maps into a hydration score.

What Is It?

Polarization describes the direction in which light waves wiggle. When light passes through aligned fibers (like collagen), polarization shifts in predictable ways.

Birefringence is the difference in refractive index along two directions. Aligned tissues like nails and skin show birefringence that depends on hydration and collagen organization.

Stacking a polarizer film on the camera and rotating it gives you a series of images that encode birefringence. A regression CNN learns the mapping from these images to measured hydration or collagen-disorder scores.

Why This Is a Good Topic

Smartphone optical diagnostics is an active research area. Hardware costs are pocket change. You will learn polarimetry, regression CNNs, and clinical-validation methodology.

Research Questions

• How does polarizer angle step size change prediction accuracy?
• What is the effect of ambient light on signal stability?
• Does birefringence imaging beat single-image RGB?
• To what extent does skin tone shift bias?
• Which body site gives the strongest hydration signal?
• How does drying time shift the birefringence map?
• What is the effect of CNN backbone on accuracy?

Basic Materials

• Smartphone with manual-exposure app.
• Two linear-polarizer film sheets.
• 3D-printed rotation mount.
• LED ring light.
• Corneometer or weight-based hydration reference.
• Informed-consent template.

Advanced Materials

• Commercial polarimetric camera.
• Clinical hydration meter.
• Dermatology mentor.
• Calibrated rotating-polarizer setup.

Software & Tools

• PyTorch: Trains the regression CNN.
• OpenCV: Aligns polarizer-rotation image stacks.
• scikit-learn: Computes calibration plots and Bland-Altman statistics.
• ImageJ: Provides ground-truth measurements.

Experiment Steps

1. Lock the polarizer rotation steps and lighting.
2. Collect matched hydration ground truth before any model runs.
3. Build a subject-wise data split.
4. Compare polarimetric input vs. RGB-only baseline.
5. Train and report calibration plots.
6. Audit fairness across skin tones and body sites.

Common Pitfalls

• Skipping ambient-light control.
• Mixing polarizer orientations between sessions.
• Training on a narrow demographic.
• Treating high R-squared as agreement without Bland-Altman.
• Ignoring scratches on the polarizer film.

What Makes This Competitive

A competitive entry collects matched hydration measurements (corneometer or weight-based) as ground truth, runs subject-wise splits, calibrates lighting, and audits fairness across skin tones. Comparing the model to a single-image baseline tells you whether polarization actually helps.

Project Variations

• Apply the same pipeline to scleroderma screening on hands.
• Replace polarizer with cross-polarized macro photography.
• Combine polarimetry with infrared imaging for a fusion model.

Learn More

• PubMed: Search polarimetric smartphone skin imaging reviews.
• NIH PubMed Central: Open-access optical-diagnostics papers.
• OpenCV documentation: Free image-stack tutorials.
• NIST Imaging Sciences: Reference standards for polarimetry.
• MIT OpenCourseWare: Course 6.815 Computational Photography.