Model-Based Iterative Reconstruction for Bright-Field Electron Tomography

Bright-Field (BF) electron tomography (ET) has been widely used in the life sciences for 3-D imaging of biological specimens. However, while BF-ET is popular in the life sciences, 3-D BF-ET imaging has been avoided in the physical sciences due to measurement anomalies from crystalline samples caused by dynamical diffraction effects such as Bragg scatter. In practice, these measurement anomalies cause undesirable artifacts in 3-D reconstructions computed using filtered back-projection (FBP). Alternatively, model-based iterative reconstruction (MBIR) is a powerful framework for tomographic reconstruction that combines a forward model for the measurement system and a prior model for the object to obtain reconstructions by minimizing a single cost function. In this paper, we present an MBIR algorithm for BF-ET reconstruction from crystalline materials that can account for the presence of anomalous measurements. We propose a new forward model for the acquisition system which accounts for the presence of anomalous measurements and combine it with a prior model for the object to obtain the MBIR cost function. We then propose a fast algorithm based on majorization-minimization to find a minimum of the corresponding cost function. Results on simulated as well as real data show that our method can dramatically improve reconstruction quality as compared to FBP and conventional MBIR without anomaly modeling.