Model Based Iterative Reconstruction

Model-based iterative reconstruction (MBIR) aims to improve the quality and efficiency of image reconstruction from incomplete or noisy data in various imaging modalities like CT, PET, and photoacoustic imaging. Current research focuses on integrating MBIR with deep learning techniques, such as incorporating deep neural networks within optimization frameworks (e.g., ADMM Plug and Play) or using pre-trained diffusion models as priors to enhance reconstruction accuracy and speed, particularly in challenging scenarios like sparse-view acquisitions. These advancements are significant because they enable higher-fidelity images with reduced scan times and radiation dose, leading to improved diagnostic capabilities and broader applicability in medical and industrial imaging.