An AI-based rapid estimation model for skin dose in radiation accidents
Objective Based on the practical demand for skin dose estimation in current nuclear and radiation accidents,this study aims to develop a rapid skin dose estimation model using neural networks.Methods Drawing on the theoretical framework of physical dose estimation,a BP neural network model based on genetic algorithms is constructed.This model utilizes conversion coefficients from particle fluence to skin-absorbed dose,as provided in the latest ICRU Report 95.The model is designed to swiftly estimate skin doses incurred during radiation accidents.Case studies are conducted to compare the skin dose estimations obtained using Monte Carlo simulation with those derived from the proposed conversion coefficient method,assessing the consistency between the two approaches.Results The GA-BP neural network model successfully predicts conversion coefficients for physical dose estimation,with an overall prediction error of 5.57％.In the case of accidental irradiation,the skin absorbed dose obtained from the rapid estimation model recommended in this article differs by 3.39％compared to the Monte Carlo simulation results,further validating the reliability and accuracy of the model.Conclusion The rapid skin dose estimation model developed in this study is essential for for accurately and swiftly estimating doses in radiation-induced skin diseases.
万方数据
维普
