Gradient-descent-based reconstruction for muon tomography based on automatic differentiation in PyTorch
Jean-Marco Alameddine, Felix Sattler, Maurice Stephan, Sarah Barnes
TL;DR
The paper addresses efficient muon tomography reconstruction under sparse muon statistics by formulating a likelihood-based approach that estimates voxel-wise λ. It maximizes the negative log-likelihood using PyTorch automatic differentiation, enabling flexible, gradient-based optimization. On simulated data, the gradient-descent reconstruction competes with traditional methods such as PoCA and ASR in both qualitative visuals and quantitative metrics, with comparable memory and runtime given mini-batch optimization. The authors also outline extensions, including priors and a Gaussian scale mixture model to better capture non-Gaussian scattering tails, highlighting practical impact for fast, accurate muon tomography.
Abstract
Muon scattering tomography is a well-established, non-invasive imaging technique using cosmic-ray muons. Simple algorithms, such as PoCA (Point of Closest Approach), are often utilized to reconstruct the volume of interest from the observed muon tracks. However, it is preferable to apply more advanced reconstruction algorithms to efficiently use the sparse muon statistics that are available. One approach is to formulate the reconstruction task as a likelihood-based problem, where the material properties of the reconstruction volume are treated as an optimization parameter. In this contribution, we present a reconstruction method based on directly maximizing the underlying likelihood using automatic differentiation within the PyTorch framework. We will introduce the general idea of this approach, and evaluate its advantages over conventional reconstruction methods. Furthermore, first reconstruction results for different scenarios will be presented, and the potential that this approach inherently provides will be discussed.