Abstract:Abstract In the context of nuclear safety production, the detailed information on three-dimensional radiation field dose rates plays a vital role in enhancing radiation protection and planning for radioactive areas. Research on inversion methods, especially under scenarios of sparse or irregular measurement points, has attracted significant attention. Currently, there are advances in inversion methods based on interpolation algorithms and neural network techniques. Interpolation algorithms are inversion methods capable of rapidly inverting and reconstructing the distribution of radiation field dose rates with limited radiation field data. These methods include grid data interpolation, finite element methods, inverse distance weighting, kriging, and radial basis function methods, and are the most widely used methods for the quick reconstruction of radiation fields at present. Based on the information of known sample points, these algorithms predict information at unsampled points by selecting appropriate analytical models to find suitable mapping relationships. They offer the advantages of simplicity and speed, but have poor capability in restoring the geometric shapes of multiple radiation sources and face significant challenges in the direct inversion and reconstruction of three-dimensional radiation fields. In some cases, sparse sampling data can lead to suboptimal inversion results. Neural network algorithms, renowned for their strong nonlinear mapping capabilities, can approximate complex functions by learning from extensive data sets. Once appropriately trained, neural networks can process incomplete or uncertain knowledge and predict outcomes based on limited data. These algorithms are commonly used in image restoration and inverse problem-solving. In recent radiation field dose rate inversion methods, neural network algorithms have gained attention due to their exceptional data fitting and predictive capabilities. In recent years, research has proposed an inversion method that combines interpolation with image restoration algorithms. This method utilizes the smoothness of high-order diffusion restoration equations. Through iterative solving of diffusion equations, it progressively restores the dose rate information of the three-dimensional radiation field, thus achieving rapid inversion of the radiation field dose rate distribution. This method has proven its superior performance in simple scenarios, but its efficacy in real and complex environments lacks verification. This paper implements the diffusion restoration equation for radiation field dose rate inversion in complex scenarios, validating its reliability in environments with multiple shields and materials, thereby expanding the method’s applicability. This advancement represents a significant step in the field of radiation protection, offering an effective way to reduce radiation-related risks by rapidly reconstructing accurate radiation fields in complex settings. Moreover, the integration of traditional interpolation methods with image restoration concepts highlights the potential of innovative hybrid solutions in addressing complex scientific challenges, providing a novel approach for radiation field reconstruction.

Research on a three-dimensional radiation field reconstruction algorithm based on an improved 3D CNN

A Calibration-Informed Deep Learning Model for Three-Dimensional Particle Reconstruction of Volumetric Particle Image Velocimetry

Automated Clinical Target Volume Delineation Using Deep 3D Neural Networks in Radiation Therapy of Non-small Cell Lung Cancer

Validation of the neural network for 3D photon radiation field reconstruction under various source distributions

3-D Radiation Field Reconstruction for a Facility With Multiple Radioactive Sources

3-D Gamma Dose Rate Reconstruction for a Radioactive Waste Processing Facility Using Sparse and Arbitrarily-Positioned Measurements

3-D Gamma Radiation Field Reconstruction Method Using Limited Measurements for Multiple Radioactive Sources

Improve 3D Cone-Beam CT Reconstruction by Slice-Wise Deep Learning

Learning 3D Gaussians for Extremely Sparse-View Cone-Beam CT Reconstruction

A dual-domain deep learning-based reconstruction method for fully 3D sparse data helical CT

Neural Network for 3D ICF Shell Reconstruction from Single Radiographs

Adaptive Incident Radiance Field Sampling and Reconstruction Using Deep Reinforcement Learning.

Research on High Sparse Sampling CT Image Reconstruction Algorithm Based on Convolutional Neural Networks

A Deep Learning Reconstruction Method for Sparsely Spaced Multidetector Helical CT Using Cascaded Dual-Domain CNN

Study of 3D Reconstruction Algorithm Used in Cosmic-Ray Muon Radiography

An optimized segmentation convolutional neural network with dynamic energy loss function for 3D reconstruction of lumbar spine MR images

3D Reconstruction of Ancient Buildings Using UAV Images and Neural Radiation Field with Depth Supervision

Reconstruction of Complex Scene Radiation Fields Based on Image Restoration Equation

Sparse-view CT Reconstruction with 3D Gaussian Volumetric Representation

Prediction of Radiation Pneumonitis with Dose Distribution: A Convolutional Neural Network (CNN) Based Model

Machine learning for detection of 3D features using sparse x-ray tomographic reconstruction