Wav_ele_t Prior-Based Reconstruction Algorithms for Magneto-Acousto-Electrical Computed Tomography

MAE-CT, a recently developing technique, offers superior imaging efficiency and quality, addressing challenges such as artifacts and prolonged scanning times in traditional methods. However, traditional MAE-CT reconstruction algorithms primarily focus on analytical and iterative reconstruction algorithms (IR), for example, filtered back projection (FBP) and simultaneous algebraic reconstruction technique (SART), both of which struggle to effectively suppress background noise and streak artifacts. In this paper, we propose two wavelet-domain-based algorithms that effectively separates noise and foreground, facilitating the reconstruction of high-quality MAE-CT images. The first algorithm is a traditional iterative wavelet regularization method, and the second is a data-driven deep learning algorithm. Prior information in the wavelet domain provides an optimal representation for pointwise singularities, which allows wavelets to successfully extract noise points in MAE-CT, thereby indirectly enabling the separation of conductivity boundary domain background noise. In addition, MAE-CT currently lacks sufficient data for training deep learning models. We have also proposed a method for generating simulated MAE-CT data based on publicly available CT datasets, and successfully trained it using the second data-driven algorithm presented in this study. The reconstruction results of the two methods on the simulation data outperforms FBP and iterative algorithms by better suppressing noise and streak artifacts, resulting in enhanced reconstruction quality.