Abstract
Sparse-view Reconstruction can be used to provide accelerated low dose CT imaging with both accelerated scan and reduced projection/back-projection calculation. Despite the rapid developments, image noise and artifacts still remain a major issue in the low dose protocol. In this paper, a deep learning based method named Improved GoogLeNet is proposed to remove streak artifacts due to projection missing in sparse-view CT reconstruction. Residual learning is used in GoogLeNet to study the artifacts of sparse-view CT reconstruction, and then subtracts the artifacts obtained by learning from the sparse reconstructed images, finally recovers a clear correction image. The intensity of reconstruction using the proposed method is very close to the full-view projective reconstructed image. The results indicate that the proposed method is practical and effective for reducing the artifacts and preserving the quality of the reconstructed image.
Highlights
A great effort has been devoted to improve sparse-view CT reconstruction in the past twenty years
The well-known filtered back-projection (FBP) reconstruction method is performed and the residual learning is used to remove artifacts generated during sparse-view reconstruction
Compared to the gold standard image, the FBP result suffered from the artifacts in a high degree
Summary
A great effort has been devoted to improve sparse-view CT reconstruction in the past twenty years. Base on the CS theory, a state of art solution, which is called as adaptive steepest descent projection onto convex sets (ASD-POCS) method[1], was invented by Sidky et al by minimizing the total variation (TV) of the desired image for CT image reconstruction from sparse projection views. To eliminating the patchy artifacts and preserving subtle structures, Liu et al proposed a total variation-stokes-projection onto convex sets (TVS-POCS) method[3] for the purpose of recovering possible missing information in the sparse-view data situation. These TV-based algorithms are successful in a number of cases, the power of the TV minimization constraint is still limited. Besides the TV-based method and its general case, a prior image-constrained compressed sensing (PICCS) method[4] and patch based nonlocal means (NLM)[5], tight wavelet frames, feature dictionary learning[6,7], low rank methods and so on, were introduced to further reduce the number of required projection views by incorporating prior images or patch information to the CS theory
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
