Teste | Abstract: 132-1 | ||||
Abstract:Intraoral radiographs are widely used to visualize the structures of the oral cavity, employing three main techniques: periapical, bitewing (interproximal), and occlusal. Each technique is distinguished by the positioning of the X-ray machine and image receptors, along with specific equipment settings. [1] Given the frequency with which patients may undergo these exams, especially in extensive dental treatments, there is a significant risk of exposure to primary radiation beams that can affect radiosensitive organs and tissues [2, 3, 4]. The objective of this study was to develop a machine learning (ML) model with deep learning using Multi-Layer Perceptron (MLP) techniques with hidden layers to predict the absorbed energy by radiosensitive organs of the face, such as the lens, salivary glands, and thyroid. The ML model was trained using the results of computational simulations performed with the Monte Carlo N-Particle (MCNP) version 6.2 code, with tube voltages of 60, 65, and 70 kVp and filtrations from 1.5 mmAl to 2.5 mmAl, which are recommended parameters for intraoral radiographs [5, 6, 7, 8]. For the computational simulations, we used a male anthropomorphic virtual phantom [9]. With the results of the computational simulations, we obtained the absorbed dose, through tally F6 [10, 11], in each studied organ. These values were applied in two scenarios: in training the ML model to predict deposited energy values and in calculating the Conversion Coefficients for Absorbed Dose (CC[D]) from the simulated and ML-predicted. The ML models for predicting the absorbed dose by organs can significantly reduce the computational cost and time required to process Monte Carlo simulations. Moreover, they can bring part of the computational simulations into practical application by medical physicists and other professionals in hospitals, using the results in patient dose planning. The CC[D] results indicated that the salivary glands, especially the sublingual glands, received the highest absorbed doses, followed by the parotid and submandibular glands. In some positions of the periapical and occlusal techniques, organs such as the eyes, lens, and thyroid also showed high CC[D] values due to the direction of the radiation beams and the positioning of the techniques. The authors would like to thank the Brazilian agencies CNPq (Grants 312160/2023-2 (L.P.N), 312124/2021-0 (A.P.P), 309675/2021-9 (W.S.S) and 406303/2022-3) and FAPEMIG (Grants APQ-04215-22, APQ-01254-23 and APQ-04348-23). Keywords: Intraoral radiographs, Radiosensitive organs, Machine learning models, Monte Carlo simulations, Absorbed dose prediction |