Utilização de algoritmos de IA na predição de dose radiológica em exames de tórax: uma ferramenta de apoio à física médica
DOI:
https://doi.org/10.29384/rbfm.2026.v20.19849001851Keywords:
Dose de entrada na pele, Radiografia de tórax, Inteligência artificial, Radioproteção, Física MédicaAbstract
Accurate prediction of Entrance Skin Dose (ESD) is essential for quality control and patient safety in chest radiography exams. This study investigated the application of artificial intelligence (AI) algorithms to estimate ESD based on technical factors such as mAs, kV, source-to-image distance (SID), and patient thickness. Data from 150 chest radiographs with physical dosimetry were analyzed. Supervised learning models, including an Artificial Neural Network (ANN) and a Random Forest Regressor (RFR), were trained and evaluated for predictive performance. The ANN achieved a mean absolute error (MAE) of 0.009 mGy and a coefficient of determination R² = 0.990, while the RFR achieved MAE of 0.020 mGy and R² = 0.827. The mean measured physical dose was 0.284 mGy, compared to predicted means of 0.274 mGy for the ANN and 0.282 mGy for the RFR. Both models showed good regression performance, with the ANN standing out in terms of accuracy. These findings highlight the potential of AI as a support tool for radiation protection and optimization of technical parameters in clinical environments.
Downloads
References
1. Bushberg JT, Seibert JA, Leidholdt EM, Boone JM. The essential physics of medical imaging. 3rd ed. Philadelphia: Lippincott Williams & Wilkins; 2012. 663-665.
2. International Atomic Energy Agency. Radiation protection and safety in medical uses of ionizing radiation. Specific Safety Guide SSG-46. Vienna: IAEA; 2018. p. 10-12.
3. Dance DR, Christofides S, Maidment ADA, McLean ID, Ng KH, eds. Diagnostic Radiology Physics: A Handbook for Teachers and Students. Vienna: International Atomic Energy Agency; 2014.
4. International Atomic Energy Agency. Radiation protection in diagnostic radiology. Safety Reports Series No. 457. Vienna: IAEA; 2018.
5. International Commission on Radiological Protection – ICRP. Recommendations of the International Commission on Radiological Protection. Annals of the ICRP, v. 34(1) (1991).
6. Agência Nacional de Vigilância Sanitária – ANVISA. Instrução normativa - IN nº 90, de 27 de maio (2021). Diário Oficial da União, nº 101, 31 de maio (2021).
7. Agência Nacional de Vigilância Sanitária (Brasil). Resolução RDC nº 611, de 9 de fevereiro de 2022. Dispõe sobre os requisitos sanitários para serviços de radiologia diagnóstica ou intervencionista. Diário Oficial da União [Internet]. 2022 fev 10 [citado 2025 out 6]; Seção 1:145–56.
8. Haykin, S. S. Neural networks and learning machines. [S.l.]: Pearson Upper Saddle River, NJ, USA, 2009. v. 3.
9. Abdo RC, Umisedo NK, Campos LT, Costa PR. Assessment of organ absorbed doses in patients undergoing chest X-ray examinations by Monte Carlo based softwares and phantom dosimetry. Radiat Meas. 2011;46(12):1981-1984. doi: 10.1016/j.radmeas.2011.07.017.
10. Yu H, Scalzo F, Zbijewski W, Sturgeon G, Tward DJ. Automated fetal dose estimation in CT using convolutional neurais networks and Monte Carlo simulations. Phys Med Biol. 2019;64(17):175005. doi:10.1088/1361-6560/ab332c.
11. Lin C, Hsieh Y, Lin Y, Hsu C, Chen J, et al. An artificial intelligence model for predicting an appropriate mAs with target exposure indicator for chest digital radiography. Sci Rep. 2025;15:1060. doi:10.1038/s41598-025-96947-y
12. Mori I, Nishio M, Sugiyama O, Onishi H. Reduction of radiation exposure in chest radiography using deep learning-based noise reduction. J Radiol Prot. 2025 Feb 3;45(1):012501. doi:10.1088/1361-6498/ad8a8b.
13. Isnawan K, Ardiansyah R, Suherman S. Identifikasi paparan radiasi X-Ray untuk keselamatan radiasi menggunakan Random Forest classification. Jurnal Teknika. 2023;8(2):75–81. doi:10.52561/teknika.v8i2.288.
14. Ferrante M, De Marco P, Rampado O, Gianusso L, Origgi D. Effective dose estimation in computed tomography by machine learning. Tomography. 2025 Jan 2;11(1):2. doi:10.3390/tomography11010002.
15. Ramlah R, Fadli M, Valerian J, Prajitno P, Sihono DSK. Random-Forest Model Prediction of Dose Distribution in Intensity-Modulated Radiation Therapy (IMRT) Planning for Lung Cancer. Iranian Journal of Medical Physics. 2023;20(5):298–304. doi:10.22038/ijmp.2022.64319.2101
16. Wambersie A, Zoetelief J, Menzel HG, Paretzke P. The ICRU (International Commission on Radiation Units and Measurements): its contribution to dosimetry in diagnostic and interventional radiology. Radiat Prot Dosimetry. 2005;117(1-3):7-12
17. Chakraborty S. Fundamentals of Machine Learning and Deep Learning. 1st ed. Cham: Springer; 2021.
Downloads
Published
How to Cite
Issue
Section
License
Copyright (c) 2026 Gustavo da Conceição, César Marques Salgado, Roberto Schirru
This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License.
The submission of original articles to the Brazilian Journal of Medical Physics implies the transfer, by the authors, of the rights of print and digital publication. Copyright for published articles remains with the author, with journal rights on first publication. Authors may only use the same results in other publications by clearly indicating this journal as the original publisher. As we are an open access journal, free use of articles in educational, scientific, non-commercial applications is allowed, as long as the source is cited.
The Brazilian Journal of Medical Physics is under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License (CC BY-NC-ND 4.0).
