Artificial Intelligence In Radio And Nuclear Chemistry: A Systematic Review Of Modern Research And Experimental Practices

Abstract

Background: Recent advances in artificial intelligence (AI) have transformed scientific research, particularly in complex and data-intensive fields such as radio and nuclear chemistry. Traditional experimental and analytical methods in nuclear sciences face significant limitations including high uncertainty levels, expensive computation, safety concerns, and scalability challenges. Integration of AI systems offers innovative solutions to enhance analytical accuracy, optimize experimental processes, and improve safety and efficiency in nuclear research settings.

Objective: This systematic review critically examines the application of artificial intelligence tools in modern research and experimental practices within radio and nuclear chemistry. The review, encompassing 152 peer-reviewed studies published between 2005 and 2024 across five databases (Web of Science, Scopus, IEEE Xplore, PubMed/MEDLINE, and Google Scholar), focuses on AI's impact on analytical accuracy, computational efficiency, experimental optimization, and safety enhancement in nuclear chemistry applications.

Methods: A systematic literature review was conducted following PRISMA 2020 guidelines. Comprehensive searches were performed across five electronic databases using Boolean operators and structured search strings (see Appendix A for complete search strategy). Studies employing AI techniques such as machine learning, deep learning, artificial neural networks, genetic algorithms, Bayesian inference, and hybrid physics-informed models were included. The screening process utilized Covidence software for efficient study selection. Data extraction and quality assessment were performed independently by two reviewers with inter-rater reliability of κ = 0.87 (substantial agreement). Due to methodological heterogeneity, both narrative synthesis and comparative quantitative analysis were employed.

Results: Analysis of 152 peer-reviewed articles revealed substantial and consistent growth in AI-based solutions across multiple nuclear chemistry domains. Deep learning models demonstrated improvements of 25-40% in radiochemical analysis and spectroscopy accuracy, with computational time reductions of up to 60% while maintaining predictive accuracy above 95%. In radiation dosimetry and environmental monitoring, AI-enhanced systems improved dose estimates by 18-35% (mean: 26.5%, 95% CI: 23.1-29.9%) and enabled real-time source localization with accuracy exceeding 90%. AI-controlled radioisotope production and separation protocols achieved yield efficiency increases of 10-25% with substantial reductions (30-50%) in experimental resource consumption. Physics-informed AI models exhibited superior robustness (20-35% improvement) and interpretability compared to purely data-driven approaches.

Conclusions: Artificial intelligence tools are revolutionizing modern radio and nuclear chemistry by delivering significant improvements in analytical accuracy, experimental efficiency, and operational safety. Integrating AI with physics-based nuclear models represents a critical step toward sustainable and intelligent nuclear research. However, challenges remain in data quality, model interpretability, and regulatory compliance. Future development should prioritize: (1) robust validation frameworks including standardized performance benchmarks and uncertainty quantification protocols; (2) explainable AI models incorporating domain-specific physics knowledge; (3) international data-sharing platforms with appropriate security safeguards implementing federated learning approaches; (4) regulatory compliance mechanisms aligned with emerging AI technologies through IAEA-led working groups; and (5) interdisciplinary training programs combining nuclear science and AI expertise to develop hybrid-skilled workforce. Establishment of collaborative initiatives and harmonized standards will be essential to ensure responsible and reliable application of AI in nuclear sciences.