Machine learning-based estimation of occupational radiation dose in interventional cardiology

Abstract

In interventional cardiology, occupational radiation exposure for medical personnel can reach high levels, underscoring the critical need for effective radiation protection and monitoring methods. This study employs machine learning algorithms to estimate radiation doses received by personnel within a virtual 3D angiography room designed to reflect realistic clinical settings. Monte Carlo simulations generated radiation data across various scenarios, accounting for personnel positions, radiation source distance, and exposure angles typical in angiography. The simulation data were used to train five machine-learning algorithms (Gradient Boosting, K-nearest neighbors, Random Forest, Linear Regression, and Decision Tree). Key findings showed that machine learning models, particularly Gradient Boosting, could effectively predict dose levels by utilizing spatial and operational parameters without requiring physical dosemeter. This study provides a framework that could streamline radiation monitoring practices, making dose assessments more accessible and efficient for routine use in clinical environments.