Predicting and Monitoring Radiation Exposure of Radiographers Using Fuzzy Preprocessing and Deep Learning

• Laouni Mahmoudi

  University Mustapha Stambouli of Mascara, Algeria LISYS Laboratory, University of Mascara

• Rochdi Bachir Bouiadjra

  Computer Science Department, University Mustapha Stambouli of Mascara LISYS Laboratory, University of Mascara

• Mohammed Salem

  Computer Science Department, University Mustapha Stambouli of Mascara LISYS Laboratory, University of Mascara

This study introduces an AI-driven approach to enhance radiation exposure monitoring for radiographers, addressing the limitations of traditional personal dosimeters. These dosimeters provide passive feedback, uncertain measurements, and lack real-time or predictive capabilities. To overcome this, fuzzy logic and deep learning are integrated. Fuzzy logic categorizes continuous variables—such as radiation dose, time spent, and distance from the source—into low, medium, and high-risk levels. Deep learning is then used to predict radiation exposure and classify these risk categories. The model achieves 98% accuracy, closely matching traditional dosimeter results while significantly improving precision and recall, outperforming the model with deep learning only. This makes radiation exposure classification more reliable. By transforming continuous data into clear, interpretable risk categories, actionable safety insights are provided to guide decision-making. Furthermore, this AI-based system shifts radiation dose management from a reactive to a proactive approach, enabling real-time monitoring and reducing reliance on dosimeters.

DOI: 10.61416/ceai.v27i3.9420

• PDF