Precision occupational lead exposure assessment through medical-informed machine learning.
Xinhao Lu1, Jie Lu2, Yuan Zhao3
1School of Cyber Science and Engineering, Southeast University, Nanjing, 211189, China; State Grid Jiangsu Electric Power Co., Ltd, Nanjing Power Supply Branch, Nanjing, 210012, China.
This study introduces a novel AI model to accurately assess occupational lead exposure, improving early detection of health risks in workers. The advanced framework enhances diagnostic recall and provides interpretable insights into lead
Area of Science:
- Environmental Health
- Occupational Health
- Toxicology
- Machine Learning
Background:
- Occupational lead exposure assessment faces challenges due to single biomarker reliance, complex interactions, and algorithmic bias.
- Existing methods struggle with accurate risk stratification and early detection of subclinical lead exposure.
- There is a critical need for advanced, interpretable tools to improve lead exposure assessment in global public health.
Purpose of the Study:
- To develop and validate a novel medical prior-informed Ant Colony Optimization-Random Forest (ACO-RF) model for precise occupational lead exposure assessment.
- To integrate lead's Absorption-Distribution-Metabolism-Excretion (ADME) toxicokinetic pathways into an AI framework for enhanced biological interpretability.
- To improve the early detection and risk stratification of lead exposure in workers, reducing missed diagnoses.
Main Methods:
- Developed a novel ACO-RF model incorporating lead's ADME pathways into its algorithmic design.
- Applied a dual-layer optimization system anchored in lead's toxicokinetic pathways.
- Validated the model on a multicenter cohort of 2867 lead-exposed workers in China.
Main Results:
- Reduced feature dimensionality by over 75% (from 79 to 19 for blood lead and 16 for urine lead).
- Achieved superior predictive performance with AUC-ROC values of 0.9646 for blood lead and 0.9319 for urine lead.
- Significantly improved recall of abnormal blood lead samples from 0.409 to 0.841, enhancing detection of subclinical exposure.
Conclusions:
- The developed ACO-RF model offers a robust and interpretable tool for early lead exposure risk stratification.
- The study establishes a generalizable, knowledge-guided machine learning framework applicable to environmental and occupational health research.
- This approach addresses limitations of current methods, paving the way for improved occupational health surveillance and management.
More Related Videos
06:00Author Spotlight: Deciphering the Long-Term Effects of Low-Level Blast Exposures in Mice
Published on: May 24, 2024
04:09Predicting Treatment Response to Image-Guided Therapies Using Machine Learning: An Example for Trans-Arterial Treatment of Hepatocellular Carcinoma
Published on: October 10, 2018