Population-Level Raman Biochemical Staging of Malaria in Human Red Blood Cells Using Interpretable Machine Learning.

Accurate malaria staging is vital for treatment decisions and monitoring of transmission. Because mature Plasmodium falciparum parasites sequester in deep vasculature, noninvasive diagnostics must target circulating ring stages and transmissible gametocyte stages. Label-free optical detection is ideal as it probes intrinsic biochemical signatures without exogenous reagents. We present a population-level Raman spectroscopy framework coupled with interpretable machine learning to define biochemical signatures of infected red blood cells. By utilizing confocal and spatially offset Raman spectroscopy (SORS) on synchronized cultures, we captured aggregate biochemical alterations associated with parasite development. SHapley Additive exPlanations (SHAP) analysis identified discriminative spectral regions, specifically hemoglobin-associated vibrational modes, and composite biochemical features. To assess translational feasibility, we demonstrated sensitivity to infection signatures beneath a skin-mimetic phantom, confirming subsurface detectability under tissue-like scattering conditions. These results provide a rigorous biochemical basis for stage-specific Raman classification, laying the groundwork for noninvasive, stage-aware diagnostics capable of identifying the infection and transmission potential.