A Pressure-Stabilized Dual-Wavelength Fingertip Optical System for Robust Non-Invasive Hemoglobin Estimation and Automated Anemia Screening

Abstract

Reliable non-invasive hemoglobin (Hb) estimation
based on dual-wavelength photoplethysmogramgery (PPG) is still
hampered by instability caused by uncontrolled contact pressure
at the fingertips, perfusion and motion. It is evident that (IEEE)
experiments have successfully shown the underlying possibility of
using dual wavelengths, however, repeatability of measurements
in real-world situations still prevents large-scale deployment. This
work presents the mechanical pressure-stabilized dual-wavelength
fingertip optical system that can directly address the sensor-tissue
interface variability without increasing the spectral complexity.
Limited movement of fingertip is ensured by a confined fingertip
housing which keeps the contact force to be within 1.6-1.9 N, in
order to reduce the variance of tissue compression and stabilize
optical path length. Synchronously modulated red (660 nm) and
infrared (940 nm) signal are subject to ambient light rejection,
motion awareness filter and normalized AC/DC ratio extraction.
A calibrated regression framework is used to relate the optical
features to the hemoglobin concentration followed by an
automated classification of anemia using the WHO threshold. In
experimental evaluation (N = 32 subjects), pressure stabilization
made inter-measurement variability differences 41.8% lower than
conventional placement. For the accuracy of estimation, the results
have been improved from RMSE = 1.42 g/dL to 0.93 g/dL and
MAE from 1.16 g/dL to 0.71 g/dL. Correlation with laboratory
reference changed from r = 0.81 to r = 0.92. The proposed
approach has proven that a mechanical standardization of the
sensor-tissue interface greatly improves repeatability and
robustness of dual-wavelength Hb estimation that is crucial for
low-cost and portable solutions for population-level screening of
anemia.