Substantial researches have shown that the wildly used statistical iterative reconstruction (SIR) methods without strong constraints, for example, the maximum likelihood estimation, could induce excessive noise into reconstructions. The noise significantly degrades image quality. In this case, the traditional method of iterating till convergence is no longer feasible. In this work, we propose a structural similarity index (SSIM) based stopping criterion for SIR. We define an indicator, referred as mSSIM, of the turning point of noise amplification based on SSIM map of reconstructed images from two adjacent iterations. The mSSIM is computed from the average of SSIM map within regions of interest (ROI). A threshold of the mSSIM is set to be the stopping criterion of iterative reconstruction. We applied this strategy to the cases of two different data noise models and iterative step sizes. Experimental tests are done on two practical datasets. Result shows that we could successfully and stably obtain images of similar quality by applying this SSIM-based stopping criterion in different cases.
Sparse-view CT imaging has been a hot topic in the medical imaging field. By decreasing the number of views, dose delivered to patients can be significantly reduced. However, sparse-view CT reconstruction is an illposed problem. Serious streaking artifacts occur if reconstructed with analytical reconstruction methods. To solve this problem, many researches have been carried out to optimize in the Bayesian framework based on compressed sensing, such as applying total variation (TV) constraint. However, TV or other regularized iterative reconstruction methods are time consuming due to iterative process needed. In this work, we proposed a method of angular resolution recovery in projection domain based on deep residual convolutional neural network (CNN) so that projections at unmeasured views can be estimated accurately. We validated our method by a disjointed data set new to trained networks. With recovered projections, reconstructed images have little streaking artifacts. Details corrupted due to sparse view are recovered. This deep learning based sinogram recovery can be generalized to more data insufficient situations.