Statistical CT reconstruction using region-aware texture preserving regularization learning from prior normal-dose CT image

摘要: In some clinical applications, prior normal-dose CT (NdCT) images are available, and the valuable textures structure features in them may be used to promote follow-up low-dose (LdCT) reconstruction. This study aims learn texture information from NdCT leverage it for LdCT image reconstruction preserve features. Specifically, proposed method first learns those patches with similar structures image, can clustered by searching context efficiently surroundings of current patch. Then utilizes redundant as a priori knowledge describe specific regions image. The advanced region-aware preserving is termed ‘RATP’. main advantage PATP that properly available adaptively characterize region-specific experiments using patient data were performed evaluate performance method. RATP demonstrated superior imaging compared filtered back projection (FBP) statistical iterative (SIR) methods Gaussian regularization, Huber regularization original regularization.