1: \begin{abstract}

2: We propose a novel Learned Alternating Minimization Algorithm (LAMA) for dual-domain sparse-view CT image reconstruction. LAMA is naturally induced by a variational model for CT reconstruction with learnable nonsmooth nonconvex regularizers, which are parameterized as composite functions of deep networks in both image and sinogram domains. To minimize the objective of the model, we incorporate the smoothing technique and residual learning architecture into the design of LAMA. We show that LAMA substantially reduces network complexity, improves memory efficiency and reconstruction accuracy, and is provably convergent for reliable reconstructions. 

3: %The alternating minimization iterations are unrolled into a sequence of parallel dual-domain neural networks. The proposed approach leverages complementary information from both domains to improve the CT image reconstruction quality while suppressing artifacts and noise and preserving structural details. 

4: Extensive numerical experiments demonstrate that LAMA outperforms existing methods by a wide margin on multiple benchmark CT datasets.  

5: %

6: % Experimental results demonstrate that the proposed algorithm outperforms state-of-the-art methods by a wide margin and is memory efficient. Additionally, our analysis shows that the algorithm converges to Clarke stationary points with proper safeguard measures. The method is interpretable as a variational model, providing explainability for the reconstruction process.

7: 

8: \keywords{Learned alternating minimization algorithm \and Convergence \and Deep networks \and Sparse-view CT reconstruction.}

9: \end{abstract}

10: