Synchrotron based sparse-view CT artifact correction with STC-UNet

Abstract

Synchrotron radiation micro-CT imaging is widely used in various scientific fields as a 3D non-destructive imaging technique with high penetration, high resolution and high contrast. Roots According to the Shannon-Nyquist theorem, sufficient projection data need to be collected to obtain high-quality CT reconstructed slices. In order to improve the temporal resolution of CT, sparse data sampling methods have been proposed. However, images reconstructed from sparse view projections often have severe streak artifacts. In this paper, we propose an artifact correction method swin transformer and convolutional U-net (STC-Unet) for synchrotron sparse CT. The network is based on the structure of U-Net, which combines the local feature extraction capability of convolutional neural network and the global feature extraction capability of Transformer in the encoder part, and reduces the artifacts introduced by the single up-sampling by using a dual up-sampling module in the decoder part. The method is applied to sparse CT experiments on synchrotron radiation metal mesh samples, and the results show that the method has good results in removing artifacts while preserving structural details. Compared with other methods, the quantitative evaluation of our proposed model is significantly improved.