面向稀疏角CT重建的投影先验引导均值回归扩散桥模型OA

In sparse-view computed tomography(CT),incomplete projection data often lead to artifacts,edge blurring,and structural distortions in the reconstructed images.To address this challenge,this study proposes a diffusion bridge model based on the Ornstein-Uhlenbeck(OU)process for the conditional completion of missing projection-domain data to enable high-quality sparse-view reconstruction.The proposed method leverages the mean-reverting property of the OU process to establish a physically consistent stochastic diffusion mechanism.The diffusion bridge constraint anchors the sampling trajectory between the reconstruction prior and sparse projections,thereby achieving high-fidelity restoration of the missing data.Furthermore,a multiscale feature fusion module in the wavelet domain is incorporated into the network architecture to effectively integrate low-frequency structural information with high-frequency texture details,thereby enhancing the capability of the model to recover fine image features.Experimental results demonstrate that under the same sparse-view conditions,the proposed method outperforms existing state-of-the-art approaches in terms of both visual quality and quantitative metrics,confirming its effectiveness for sparse-view CT reconstruction tasks.