Photon

Team 15

Team Name: Tech Support

Team Members: Dhruv Bisla, Dev Mehra, Dice Jashnani, Keshav Badrinath, Kaival Shah, William Chen

Project Name: Photon

Tracks: Healthcare

Inspiration

We’re a team of hackers who find meaningful solutions to everyday problems. At MakeMIT, we had to combine this passion for building and breaking things with a meaningful problem that is near to our heart (literally).

Circulatory disorders affect hundreds of millions of people, yet the tools used to detect and monitor them are often expensive, bulky, or limited to snapshots in time. Subtle changes in blood flow can go unnoticed until they escalate into pain, tissue damage, or life-altering complications. We saw an opportunity to rethink how perfusion is measured.

These issues are especially prevalent in rural communities that have limited resources and capacity to train healthcare workers from scratch. One of the most common but crucial procedures that most people go through is getting their blood drawn. While this is seemingly routine, it requires precision and a good amount of intuition that is hard to build up without proper exposure; even then, skin tone and body fat can make routine blood work challenging for some patients. We wanted to find a way to simplify this process as much as possible in a low-cost, effective manner. Current existing solutions for vein finding are at least $5,000, which we realized was unnecessarily expensive and impractical for underserved communities.

What it does

Photon functions as a non‑contact perfusion scanner positioned roughly 3 inches above the arm, allowing it to measure blood‑flow signals without touching the skin. In doing so, Photon can map the location of every vein across the body in just a matter of seconds.

This is done by emitting infrared light that penetrates the tissue, while a sensor captures subtle fluctuations in the reflected light over time. By analyzing these variations, the device reconstructs a spatial map of blood flow, highlighting differences across the scanned area. This works because hemoglobin absorbs near-infrared light more strongly than surrounding skin tissue, causing superficial veins to appear as darker structures in the raw stream.

By mapping veins across a patient’s body, Photon can make vein visualization accessible to rural communities that may never receive expensive clinical support. Longitudinal tracking further allows clinicians to monitor treatment response or disease progression, turning a rapid optical scan into a tool for triage, procedural guidance (e.g., IV placement), and early detection of circulatory compromise before macroscopic symptoms appear.

How we built it

We started with testing the constituent electronic components. Having not worked with near infrared LEDs before, we decided to build a setup to test them with a USB-C PD based DC power input and constant current driver. We noticed that, while IR isn’t in the visible spectrum of light, our phone cameras were a quick sanity check to debug LED control. Our next step was to prototype the electronics setup for the device. We used 3 constant current LED drivers wired in parallel, powered by a USB-C PD breakout that we set to negotiate for 12V. Our next steps were to build a proof of concept using a ring of the NIR LEDs in order to illuminate the region of interest of the camera. Building a laser cut system with variable spacing allowed us to empirically tune parameters like lens distance to sensor for focusing and camera separation from the skin for the best signal to noise ratio. With some verification of the forward voltage and current we saw experimentally, we determined that an appropriate NIR illumination setup with our materials consisted of 3 sets of 6 series LEDs, each driven by its own constant current driver (appropriate constant current was achieved by selecting a suitable 330 Ohm resistor for the constant current supply given our LED IV curve.

Having verified each of the components responsible for the function of our device, we moved to soldering the wires for a more robust connection and began designing a case to house the overall camera assembly. We opted for a purely laser cut design for our device given the ability to far more quickly make fixes to the design when compared to 3d printing (which, for the volume of our device alone, would have taken more than 3 times the imposed 3 hour limit). The laser cut assembly was designed to use tab and slot connections made with an interference fit such that parts could be sufficiently held in place with a little super glue. The core features integrated into the mechanical design include an ergonomic handle to grasp the device, a mount for the raspberry pi, and mounts for the camera and LED ring at the determined distances away from the skin.

To enhance vessel visibility, we applied Contrast Limited Adaptive Histogram Equalization (CLAHE), which performs localized histogram normalization across small image tiles. This method improves ridge and edge definition while preventing noise amplification that would occur with global histogram equalization.

After contrast enhancement, threshold-based ridge extraction and frame smoothing were used to generate a real-time perfusion map that emphasizes persistent vascular structures as the device is scanned across the skin.

Challenges we ran into

Optical focusing on skin
Achieving consistent focus at a fixed working distance was difficult due to curvature, motion, and varying reflectance.
Signal processing pipeline
- Extracting a stable physiological signal from the raw stream required iterative filtering and temporal smoothing.
Noise reduction in vein mapping
- Applying CLAHE significantly improved local contrast, but required careful tuning to avoid amplifying sensor noise.
Vein isolation limitations in visible light
- Reliable vessel extraction was only possible after switching to near-infrared illumination, aligning with findings in prior literature.

Accomplishments that we're proud of

Getting vein detection to work in under 24 hours, making notable progress toward low-cost medical hardware.

What we learned

How to develop a manual near-infrared light setup with the OV9281 image sensor
How to compartmentalize embedded system components
Rapid CAD iteration
Different image processing and noise reduction strategies in OpenCV

What's next for Photon

A stand to free up hands
- Allows for easier intravenous operations
- Makes results even more consistent and dependable
Improved software analysis
- Provides deeper insights into blood clot formation
Better internal electronics assembly
More streamlined power internals
Perhaps having a tool to mark the location of the vein for better whole-body vein mapping