A dataset-free deep learning method for low-dose CT image reconstruction

Low-dose CT (LDCT) imaging attracted a considerable interest for the reduction of object's exposure to X-ray radiation. In recent years, supervised deep learning (DL) has been extensively studied LDCT image reconstruction, which trains network over dataset containing many pairs normal-dose and low-dose images. However, challenge on collecting such in clinical setup limits application supervised-learning-based methods reconstruction practice. Aiming at addressing challenges raised by collection training dataset, this paper proposed unsupervised method does not require any external data. The is built re-parametrization technique Bayesian inference via with random weights, combined additional total variational~(TV) regularization. experiments show that noticeably outperforms existing dataset-free test