São Paulo, Brazil / Alabama, United States

The project focuses on the development of a physics-informed neural network (PINN) capable of reconstructing neutron source images in extremely low-count environments, scenarios where detectors register only a handful of neutron events and conventional imaging techniques struggle or fail entirely. By embedding detector response and imaging physics directly into the neural network, the approach enables reliable source reconstruction from sparse, noisy data while dramatically reducing computation time.

Advancing Neutron Imaging Under Real-World Constraints

Neutron imaging plays a critical role in nuclear safety, security, nonproliferation, and emergency response. Neutrons can penetrate shielding materials, making them essential for detecting and locating special nuclear materials in complex environments. However, many field and laboratory systems produce images of limited quality due to inherently low neutron counts.

This research builds upon prior experimental systems designed for low-signal conditions and replaces slow, iterative reconstruction algorithms with a physics-informed neural network that integrates detector physics directly into its loss function. The result is faster, clearer imaging even when data is severely limited. Validation will be performed using high-fidelity nuclear physics simulations based on Geometry and Tracking v4 (GEANT4), establishing a benchmark for future neutron imaging systems.

Practical Impact

The technology under development has direct implications for real-world scenarios such as cargo screening at ports of entry, rapid assessment of suspicious radiation sources, and emergency response operations. By reducing reconstruction time and increasing reliability, the project moves neutron imaging closer to operational deployment rather than remaining confined to laboratory demonstrations.

The project will deliver:

An open-source PINN-based neutron image reconstruction framework

A benchmark simulation dataset for the research community

Demonstrated performance advantages over traditional reconstruction methods in sparse-data regimes

Goals of the Research

The primary goal is to enable accurate localization and reconstruction of neutron source distributions from extremely low detector counts. Specific objectives include:

Designing and training a physics-informed neural network that embeds detector-response physics directly into its optimization process

Demonstrating superior performance over conventional methods when data is sparse

Validating results using high-fidelity Monte Carlo–simulated detector data

Releasing an open-source reconstruction tool to accelerate further research

About the Research Team

The project is led by Adam Glick, Ph.D., nuclear physicist and software engineer, alongside a multidisciplinary team with deep expertise in nuclear physics, electrical engineering, embedded systems, and real-time signal processing.

Adam Glick, Ph.D. is a nuclear physicist and software engineer with more than a decade of experience in high-performance software development, AI, and Monte Carlo simulations for nuclear imaging and radiation therapy. His work spans national laboratories, industry, and healthcare research, with publications in leading peer-reviewed journals.

Miles O’Brien is an electrical engineer with extensive experience in radar algorithms, large-scale code modernization, multithreading, and real-time systems, much of it in classified environments.

Mustapha Saad is a software engineer specializing in embedded systems, computer vision, and radiation detection, with professional experience across semiconductor fabrication, cloud computing, and command-and-control systems.

About Corporação Saulo Neto

Corporação Saulo Neto is a privately funded Brazilian initiative focused on experimental research and development in science and technology. Operating independently of government agencies, the organization supports interdisciplinary projects that combine rigorous scientific methodology with advanced computation, artificial intelligence, and applied engineering. By backing international research efforts such as this project, Corporação Saulo Neto aims to accelerate innovation with tangible impact on security, safety, and scientific progress.

Statement of Support

Through this support, Corporação Saulo Neto reaffirms its commitment to advancing high-impact scientific research beyond national borders, fostering collaboration between independent laboratories and private R&D initiatives, and promoting open scientific tools that benefit the global research community.