Machine learning and neural network approaches for enhanced measuring and prediction of radiation doses

A bstract Precise radiation dose measuring and prediction are crucial for ensuring safe and responsible use of ionizing radiation in various fields. However, traditional methods face certain limitations. Machine learning offers potential benefits such as improved accuracy, efficiency and real-time capability. A comprehensive literature review was conducted searching relevant databases. Studies applying machine learning for tasks such as dose detection, measurement, forecasting, simulation and organ dose prediction were analyzed. Techniques including classification, regression, clustering, time series analysis and simulation/generative models were assessed. Machine learning has demonstrated success in applications such as radiation source detection, continuous dose measurement, grouping dose profiles, time series forecasting, scenario-based prediction and organ dose estimation. Integration with medical imaging and radiation therapy has optimized treatment. Diverse domains from nuclear facilities to emergency response have benefited. Machine learning shows tremendous potential to revolutionize radiation sciences through personalized, real-time and optimized dose management with higher accuracy and efficiency. As data and computing power grow, its impact is poised to further transform diverse technical and clinical areas of radiation protection, monitoring and prediction.