Introduction
Space radiation represents one of the most critical health challenges for astronauts during long-duration missions. Its effects stem from complex atomic and subatomic interactions involving heavy ions and high-energy photons that induce biological damage in tissues. Traditional computational dosimetry and Monte Carlo simulations have provided valuable insight, but their predictive capacity remains limited in highly variable radiation environments. Artificial Intelligence (AI) offers a transformative approach to modeling these atomic-scale processes and predicting their biomedical consequences with greater precision.
Methods
This systematic review followed PRISMA guidelines and analyzed studies indexed in PubMed, Scopus, IEEE Xplore, and NASA ADS from 2010 to 2025. The inclusion criteria focused on research employing machine learning (ML) or deep learning (DL) algorithms for atomic-level radiation modeling, dose prediction, or biological risk estimation. Extracted data included algorithm type, dataset source, atomic modeling scale, and biomedical applications.
Results and Discussion
Across 68 eligible studies, AI-based models outperformed traditional analytical or Monte Carlo methods in radiation prediction accuracy by an average of 25-40%. Neural networks and ensemble learning approaches showed superior performance in correlating atomic interaction data with biomarkers of DNA damage, oxidative stress, and neurocognitive decline. Hybrid AI frameworks integrating atomic collision data and biomedical endpoints demonstrated promising applications for astronaut health risk prediction and adaptive shielding design.
Conclusions
AI-driven modeling of atomic and space radiation interactions is redefining the landscape of space biomedicine. Integrating atomic data with biomedical outcomes enables more accurate and individualized radiation risk assessment, essential for future Moon and Mars missions. Future research should prioritize standardized datasets and explainable AI models to bridge atomic physics and biomedical prediction systems.
