Eigenpulse: Eliminating Demographic Bias in Pulse Oximetry and Remote PPG from First Principles

Overview

When my father was hospitalized, overworked staff struggled to check his vitals regularly. That experience led me to remote heart-rate monitoring, then to a deeper problem: optical heart-rate systems falsely give normal readings on darker skin. Black patients experience three times the rate of undetected hypoxemia, contributing to 41% increased mortality. This bias was first documented in 1990, and despite decades of research, it remains unresolved. Most approaches focused on improving signal-to-noise ratio, but I returned to the underlying physics. I discovered a second pulsatile component at cardiac frequency that had been overlooked, and proved it is a dominant cause of demographic error. This led to three innovations: a method that improves equity by removing this signal component, a framework that separates coupled vascular-signals, and a three-wavelength oximetry system reducing pulse-oximetry bias from 2.3% to under 0.15%, without training data. The problem was never unsolvable; it was unexamined.

Awards (6)

• Best in Fair
• Platinum Award
• Young Scientist Award
• Challenge Award
• Gold Medal
• Selected for CWSF 2026

Competition history

• CWSF 2026 Health & Wellness

Related projects

Closest projects by meaning, across every fair and year in the corpus.