An unsupervised reconstruction method for low-dose CT using deep generative regularization prior

Low-dose CT imaging requires reconstruction from noisy indirect measurements which can be defined as an ill-posed linear inverse problem. In addition to conventional FBP method in imaging, recent compressed sensing based methods exploit handcrafted priors are mostly simplistic and hard determine. More recently, deep learning (DL) have become popular medical field. DL try learn a function that maps low-dose images normal-dose images. Although the results of these promising, their success depends on availability high-quality massive datasets. this study, we proposed does not require any training data or process. Our exploits such approach convolutional neural networks (CNNs) generate patterns easier than noise, therefore randomly initialized generative suitable used regularizing reconstruction. experiments, is implemented with different loss variants. Both analytical phantoms real-world views. Conventional method, iterative (SART), TV regularized SART comparisons. We demonstrated our variants outperforms other both qualitatively quantitatively.