Medical Radiology and Radiation Safety. 2025. Vol. 70. № 4

DOI:10.33266/1024-6177-2025-70-4-46-54

V.G. Barchukov, A.A. Bolotov, Y.N. Zhirnov, A.S. Samoylov, S.M. Shinkarev,
I.B. Ushakov, I.K. Tesnov, A.S. Galuzin, D.A. Kudinova, V.U. Lizunov

Using Artificial Intelligence Technologies for Radiation Protection 
during Decommissioning of Radiation and Nuclear Facilities

A.I. Burnazyan Federal Medical Biophysical Center, Moscow, Russia

ABSTRACT

Background: The decommissioning of radiation-hazardous facilities (RHF) is a complex process that requires compliance with legislative and regulatory requirements. Effective document management and continuous personnel training are essential for ensuring safety and regulatory adherence. Artificial Intelligence (AI) can significantly simplify and automate documentation processing and management, reducing staff workload and minimizing errors. Additionally, AI helps ensure regulatory compliance by automatically tracking changes and maintaining adherence to standards. Furthermore, AI can analyze large volumes of data, identify potential risks, and propose optimal solutions based on predictive analytics.

Purpose: To develop an AI-based service capable of supporting a comprehensive and informed dialogue on RHF decommissioning. To achieve this, we selected a natural language processing (NLP) model based on Keras. A dataset was created for training the model, consisting of five key regulatory documents on radiation safety. The documents were divided into separate contexts, with experts formulating corresponding questions and answers. In total, 429 contexts were processed, and 6,405 questions and answers were generated.

Results: The model was tested in a specially developed application similar to ChatGPT, designed to help specialists find answers to questions arising during the decommissioning process. Additionally, a dynamic knowledge base update feature was implemented, allowing for real-time adjustments to regulatory documentation changes. The developed system demonstrated high accuracy in answering questions related to regulatory aspects of decommissioning. Machine learning algorithms trained on our dataset for text processing and interpretation proved effective in recognizing and handling user queries. The system was tested in various scenarios, including internal Keras model evaluations and test questions not included in the training dataset.

Conclusion: The obtained results confirmed the potential of AI technologies in managing RHF decommissioning processes. Furthermore, tests conducted on real-world data helped identify key areas for further system improvement and functional expansion.

Keywords: Artificial intelligence, decommissioning of radiation and nuclear facilities, regulatory documents, radiation safety, Natural Language Processing (NLP), Keras

For citation: Barchukov VG, Bolotov AA, Zhirnov YN, Samoylov AS, Shinkarev SM, Ushakov IB, Tesnov IK, Galuzin AS, Kudinova DA, Lizunov VU. Using Artificial Intelligence Technologies for Radiation Protection during Decommissioning of Radiation and Nuclear Facilities. Medical Radiology and Radiation Safety. 2025;70(4):46–54. (In Russian). DOI:10.33266/1024-6177-2025-70-4-46-54

References

1. Devlin J., Chang M.W., Lee K., Toutanova K. BERT: Pre-Training of Deep Bidirectional Transformers for Language Understanding. arXiv preprint arXiv:1810.04805.

2. Sun C., Qiu X., Xu Y., Huang X. How to Fine-Tune BERT for Text Classification? China National Conference on Chinese Computational Linguistics. Springer, Cham. P. 194-206.

3. Wolf T., Debut L., Sanh V., Chaumond J., Delangue C., Moi A., Rush A.M. Transformers: State-of-the-Art Natural Language Processing. Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing: System Demonstrations. P. 38-45.

4. Radford A., Narasimhan K., Salimans T., Sutskever I. Improving Language Understanding by Generative Pre-Training. OpenAI. 2018.

5. Floridi L., Chiriatti M. GPT-3: Its Nature, Scope, Limits, and Consequences. Minds and Machines. 2020;30;4:681-694.

6. Liu Y., Ott M., Goyal N., Du J., Joshi M., Chen D., Stoyanov V. RoBERTa: A Robustly Optimized BERT Pretraining Approach.  arXiv. 2019 preprint arXiv:1907.11692.

7. Gulli A., Sujit P. Deep Learning with Keras: Implementing Deep Learning Models and Neural Networks with the Power of Python. Birmingham, Packt, 2017. 318 p.

8. Anshik. AI for Healthcare: Keras and TensorFlow. AI and Machine Learning for Healthcare. New Delhi, 2021. 381 p.

 PDF (RUS) Full-text article (in Russian)

Conflict of interest. The authors declare no conflict of interest.

Financing. The study had no sponsorship.

Contribution. Article was prepared with equal participation of the authors.

Article received: 20.03.2025. Accepted for publication: 25.04.2025.