Optimisation of PET Image Reconstruction: Use of Neural Networks for Scatter Estimation

摘要: This study aims to exploit neural networks to develop innovative methods for estimating diffusion, improving the accuracy/computation time ratio in the PET (Positron Emission Tomography) image reconstruction process. The methodology is based on Monte-Carlo simulations, which have been used to generate a large set of realistic PET data, essential for training neural networks. Detailed modelling of tomographs and analytical phantoms with a variety of characteristics was used to create a substantial learning base. The results show a significant improvement in the accuracy of reconstructions while significantly reducing computing time, demonstrating the potential of neural network-based approaches for optimising reconstruction processes in medical imaging.