Near-Infrared Vein Finder

By /

Near-Infrared Vein Finder

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

Nurses sometimes miss veins three or four times before getting a needle in. Hospitals buy 4000-dollar NIR vein finders to make this easier. The same physics works with 5 dollars of 940 nm LEDs and a phone whose IR filter you removed. A small CNN trained on the VEINS-VITAL dataset cleans up the images.

What Is It?

Veins absorb near-infrared light more than surrounding tissue. Shining 940 nm light on a forearm and imaging with an IR-sensitive camera makes veins appear as dark stripes.

Most phone cameras have an IR-cut filter inside. Removing it on a cheap secondary phone lets the sensor pick up NIR. A simple LED ring provides the light source.

A CNN trained on public NIR vein datasets enhances the raw image, sharpening vein contrast and segmenting branches. The output guides venipuncture targets.

Why This Is a Good Topic

Affordable vein imaging is real-world useful and approachable. The hardware story is striking. You will learn NIR optics, camera modification, and image-segmentation CNNs.

Research Questions

  • How does LED wavelength change vein contrast?
  • What is the effect of subject skin tone on segmentation?
  • Does the CNN beat unsharp masking as a baseline?
  • To what extent does arm temperature shift contrast?
  • Which LED array geometry maximizes uniform illumination?
  • How does camera exposure affect SNR?
  • What is the effect of training-set size on cross-subject accuracy?

Basic Materials

  • 940 nm IR LEDs and resistors.
  • Used phone with IR filter removed.
  • 3D-printed LED ring mount.
  • Power source for LEDs.
  • Tripod and arm rest.
  • Informed-consent template.

Advanced Materials

  • Clinical NIR vein finder for ground-truth comparison.
  • High-quantum-efficiency NIR camera.
  • Calibrated phantom with embedded tubing.
  • Cloud GPU.

Software & Tools

  • PyTorch: Trains the segmentation CNN.
  • OpenCV: Preprocesses NIR images.
  • Roboflow: Manages dataset labels.
  • Matplotlib: Plots contrast metrics.

Experiment Steps

  1. Confirm IR filter removal and document the camera mod.
  2. Lock LED geometry and exposure.
  3. Build a subject-wise data split.
  4. Train and validate the segmentation CNN.
  5. Report contrast and segmentation metrics.
  6. Audit fairness across skin tones.

Common Pitfalls

  • Letting LED current drift, changing illumination.
  • Training on one skin tone.
  • Treating visual nicety as quantitative improvement.
  • Ignoring ambient IR from sunlight.
  • Comparing only against trivial baselines.

What Makes This Competitive

Run subject-wise data splits, report contrast metrics not just visual examples, and audit fairness across skin tones and arm sizes. Comparing the CNN output to manual annotations on held-out subjects validates the enhancement.

Project Variations

  • Add real-time vein overlay via AR.
  • Switch to multispectral imaging with two NIR bands.
  • Build a child-arm phantom with tubing for safe practice.

Learn More

  • PubMed: Search NIR vein imaging review.
  • NIH PubMed Central: Open-access venipuncture assistance papers.
  • OpenCV documentation: Free imaging tutorials.
  • NIST Imaging Sciences: Reference contrast metrics.
  • MIT OpenCourseWare: Course 6.815 Computational Photography.
Shopping Cart