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

Area of Science:

• Biomedical Optics
• Spectroscopy
• Machine Learning

Background:

• Accurate malaria staging is crucial for effective treatment and transmission monitoring.
• Mature Plasmodium falciparum parasites sequester, necessitating noninvasive diagnostics targeting circulating ring and gametocyte stages.
• Label-free optical detection offers reagent-free analysis of intrinsic biochemical signatures.

Purpose of the Study:

• To develop a population-level Raman spectroscopy framework for defining biochemical signatures of malaria-infected red blood cells.
• To utilize interpretable machine learning for stage-specific parasite classification.
• To assess the feasibility of noninvasive, subsurface detection of malaria infection signatures.

Main Methods:

• Confocal and spatially offset Raman spectroscopy (SORS) were applied to synchronized Plasmodium falciparum cultures.
• Interpretable machine learning, specifically SHapley Additive exPlanations (SHAP), was used to identify discriminative spectral features.
• A skin-mimetic phantom was used to evaluate subsurface detection capabilities.

Main Results:

• Aggregate biochemical alterations associated with parasite development were captured.
• SHAP analysis identified key spectral regions, including hemoglobin-associated vibrational modes, as discriminative features.
• Subsurface detection of infection signatures was demonstrated through a tissue-mimetic phantom, confirming detectability under scattering conditions.

Conclusions:

• A rigorous biochemical basis for stage-specific Raman classification of malaria parasites was established.
• The framework provides a foundation for developing noninvasive, stage-aware diagnostics.
• This approach has the potential to identify malaria infection and assess transmission potential.