Application of the crAIRsis AI-Based Framework for the Analysis of PCB-170 in human breast milk

Keywords:
Human Biomonitoring, Polychlorinated Biphenyls, Machine Learning, Metaheuristics, Explainable Artificial Intelligence

Abstract:
Breast milk is a reliable, non-invasive matrix for monitoring internal exposure to polychlorinated biphenyls (PCBs) and organochlorine pesticides (OCPs), particularly in vulnerable populations such as infants. Ongoing monitoring studies underscore the need for a deeper understanding of the distribution and health impacts of these persistent organic pollutants. Although artificial intelligence (AI) has been widely applied across scientific disciplines, its use in environmental exposure analysis, particularly in biological matrices like breast milk, remains limited. This study investigates the distribution of PCB-170, a highly chlorinated and toxicologically relevant PCB congener, and identifies key predictive factors using an advanced AI-based framework. The analysis was performed using the crAIRsis platform, which integrates ensemble machine learning algorithms, metaheuristic optimisation, and explainable AI methods such as Shapley additive explanations (SHAP) and Shapley additive global importance (SAGE). This approach enables the modelling of complex, nonlinear relationships between variables. Breast milk samples from 186 mothers in Zadar, Croatia, were analysed for 17 PCB congeners and 7 OCPs. The most influential predictors of PCB-170 levels were PCB-180, PCB-153, and PCB138, indicating strong co-behaviour and likely shared exposure pathways. These congeners showed relative SHAP impacts ranging from -40% to over 60%. Demographic variables, including maternal age and birth order, had minimal influence, with SHAP impacts below 10%. The study demonstrates the dominant role of higher-chlorinated PCBs in shaping internal burdens and highlights the value of explainable AI in environmental health research. The crAIRsis framework offers a robust, transferable methodology for human biomonitoring and evidence-based exposure assessment.

CITATION:

IEEE format

Copy to clipboard

T. Bezdan, G. Jovanović, A. Stojić, S. Herceg Romanić, M. Perišić, “Application of the crAIRsis AI-Based Framework for the Analysis of PCB-170 in human breast milk,” in Sinteza 2025 - International Scientific Conference on Information Technology, Computer Science, and Data Science, Belgrade, Singidunum University, Serbia, 2025, pp. 221-227. doi:10.15308/Sinteza-2025-221-227

APA format

Copy to clipboard

Bezdan, T., Jovanović, G., Stojić, A., Herceg Romanić, S., Perišić, M. (2025). Application of the crAIRsis AI-Based Framework for the Analysis of PCB-170 in human breast milk. Paper presented at Sinteza 2025 - International Scientific Conference on Information Technology, Computer Science, and Data Science. doi:10.15308/Sinteza-2025-221-227

BibTeX format
Download

RefWorks Tagged format
Download