CBCTformer: Transformer Model for Sparse-View CBCT Artifact Reduction in Image-Guided Radiotherapy

This study proposes a Transformer-based image domain model, named CBCTformer, for artifact reduction in sparse-view cone-beam computed tomography (CBCT) used in image-guided radiotherapy (IGRT). The CBCTformer model is built on a Swin Transformer U-Net architecture, comprising both encoder and decoder components. The model was trained, validated, and tested on CBCT data from 163, 30, and 30 real patients, respectively. Quantitative metrics including root-mean-square error (RMSE), peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and feature similarity (FSIM) were used to evaluate the model's performance. The CBCTformer model was compared with four state-of-art (SOTA) convolutional neural network (CNN)-based models on the 1/8 sparse-view reconstruction task. On the test set, the CBCTformer model achieved averaged RMSE, PSNR, SSIM and FSIM values of 0.001098, 35.67 dB, 0.8714 and 0.97, respectively, outperforming all the four SOTA CNN models across all metrics. In conclusion, the proposed CBCTformer model effectively reduces artifacts in sparse-view CBCT images, thereby improving CBCT image quality in IGRT.