Iterative Reconstruction for Multi-Modal Neutron Tomography

We describe a unified framework for model-based iterative 3D reconstruction of multi-modal neutron transmission, hydrogen-scatter, and induced-fission images from low resolution data recorded using 14.1MeV neutrons and the associated-particle imaging (API) technique. The framework, which was developed to facilitate use in challenging field-deployment scenarios, is centered around physics-based system models and a total variation constrained implementation of the Simultaneous Iterative Reconstruction Technique (SIRT). Modified to solve a statistically weighted least squares problem, the SIRT algorithm is accelerated using ordered subsets and Nesterov’s momentum for which we derive a near-optimal value of the governing Lipschitz constant. The approach enables reconstruction of images that are high resolution compared to the acquired data and is robust to both limited statistics and a limited number of projection angles. Moreover, the framework is fast enough to be practical. Example images are provided that demonstrate both the ability to perform fast-neutron imaging of high-atomic-number materials with low radiation dose and the benefit of multi-modal neutron imaging to identify key materials.