Back to projects
Edge Medical AIComputer Vision

Urinalysis

2024 — 2025 · Deployed in pilot
10
Test strip parameters read
<2s
Time to result
No lab
reader required

Full write-up coming soon — detailed content for this project is being finalized.

Overview

Standard urinalysis test strips are cheap and widely available, but reading them accurately still usually requires a dedicated lab reader. This project reads strip results directly from a smartphone photo, making the same test usable anywhere a phone camera is.

Approach

The core challenge is color constancy: a strip photographed under fluorescent light, sunlight, or a phone's flash produces very different raw colors for the same result. The model corrects for ambient lighting using a reference patch on the strip itself, then classifies each pad against a calibrated color reference.

Results

Across a validation set spanning multiple lighting conditions and phone models, strip readings matched lab-reader results closely enough to support triage-level decisions.

Point-of-careComputer vision