Training a smartphone-based oximetry algorithm using low-cost easy-to-prepare tissue-simulating phantoms

Conventional pulse oximetry exhibits variable accuracy across patient populations. A smartphone-based alternative, if calibrated across diverse tissue properties, could improve diagnostic equity. We present low-cost, tunable, biomimetic phantoms simulating oxy- and deoxyhemoglobin spectra using magenta and purple pigments. These phantoms emulate hemoglobin’s bimodal peaks (~550 nm) and exhibit a smaller 625 nm peak indicative of methemoglobin. RGB image analysis showed a strong R² = 0.98 correlation between red-to-blue ratios and pigment concentration, supporting algorithm training for oxygenation assessment. The heightened 625 nm peak sensitivity suggests hypoxia modeling. These phantoms enable optical property quantification and are ideal for research and education in low-resource environments.