Patient’s exposure to radiation has come under scrutiny as X-ray computed tomography (CT) has become more used in clinical diagnosis. However, doing so will produce server noise and undermine radiologists' trust in their diagnostic abilities. In order to enhance picture quality, it is necessary to create progressive poor-dose CT (PDCT) image reconstruction techniques. The last two years have seen remarkable progress in noise reduction for PDCT pictures using deep learning-based methods. However, acquiring well-matched datasets involves repeated scans, which in turn increases radiation exposure, and hence most contemporary deep learning-based algorithms need the paired training dataset in which the PDCT pictures correspond to the normal-dose CT (NDCT) images exactly. As a result, high-quality matched datasets are hard to come by. This study presents an unpaired PDCT image denoising network that is based in cycle-generative adversarial networks (CycleGAN) employing previous image information as a solution for this issue as this kind of network doesn't require an individual training data set. A key technique for unpaired image-to-image translationis cycle loss is a promising method of mapping the distribution of PDCT to NDCT by using this method. Furthermore, the prior information gathered from the results of the pre-processing by using the PDCT image is integrated into the network in order to supervise the evolution of content, which ensures that there is a perfect match with the other. The technique we have proposed not only reduces image noise, but also ensures that important information remains intact using a distribution map generated from cyclic loss as well as content supervision derived from loss of images in the past. The efficacy of our suggested technique was evaluated using real-data tests. If compared to the initial CycleGAN with no prior knowledge it is found that the SSIM increases and the peak signal-tonoise-ratio (PSNR) is improved by 3-dB. The numerical and visual evaluations of the actual PDCT experiment prove the effectiveness of the proposed method

tomography, CT Diagnose, image-to-image translation, structural similarity, peak signalto-noise ratio