Tenzing H. Y. Joshi

In this third part of a multi-paper series, we present quantitative image reconstruction results from aerial measurements of eight different surrogate distributed gamma-ray sources on flat terrain. We show that our quantitative imaging methods can accurately reconstruct the expected shapes, and, after appropriate calibration, the absolute activity of the distributed sources. We conduct several studies of imaging performance versus various measurement and reconstruction parameters, including detector altitude and raster pass spacing, data and modeling fidelity, and regularization type and strength. The imaging quality performance is quantified using various quantitative image quality metrics. Our results confirm the utility of point source arrays as surrogates for truly distributed radiological sources, and advance the quantitative capabilities of Scene Data Fusion gamma-ray imaging methods.

Open, realistic datasets are essential for developing and benchmarking radiation detection algorithms, yet they remain scarce. We introduce the Radiological Anomaly Detection and Identification (RADAI) dataset, a large-scale synthetic resource that integrates high-fidelity Monte Carlo simulations with realistic urban scenarios to capture both background variability and source signatures. RADAI models construction-material NORM, people and vehicles, urban clutter, and dynamic environmental effects such as cosmic-ray and rain-induced transients, and it provides list-mode detector data with motion and response modeling suitable for algorithm training and evaluation. We publicly release three complementary datasets for this purpose (training, developer, and testing) together with an online scoring portal for standardized performance assessment and an open software toolkit that supports data access, augmentation, model development, and evaluation. These resources enable reproducible comparisons across methods and promote rigorous studies at the scale required by contemporary machine learning. By grounding algorithm development in realistic, well-documented conditions, RADAI supports progress toward more robust detection, identification, and localization in complex urban environments.

Sparse static detector networks in urban environments can be used in efforts to detect illicit radioactive sources, such as stolen nuclear material or radioactive "dirty bombs". We use detailed simulations to evaluate multiple configurations of detector networks and their ability to detect sources moving through a 6 × 6km ² area of downtown Chicago. A detector network’s probability of detecting a source increases with detector density but can also be increased with strategic node placement. We show that the ability to fuse correlated data from a source-carrying vehicle passing by multiple detectors can significantly contribute to the overall detection probability. In this paper we distinguish static sensor deployments operated as networks able to correlate signals between sensors, from deployments operated as arrays where each sensor is operated individually. In particular, we show that additional visual attributes of source-carrying vehicles, such as vehicle color and make, can greatly improve the ability of a detector network to detect illicit sources.

In this first part of a multi-paper series, we demonstrate a method for using arrays of point sources to emulate continuous distributed gamma-ray sources when measured from a standoff of at least several meters. The method relies on the Poisson deviance statistic to test whether the array source “looks like” its continuous analogue when measured by a particular gamma-ray detector moving through 3D space on a particular trajectory. This point-source method offers significant advantages over truly distributed sources such as powders, solutions, or aerosols; notably, arrays of sealed point sources are safer to both personnel and the environment, and are more easily deployed, reconfigured, ground-truthed, and removed. We use this Poisson deviance metric to design eight different mock distributed sources, ranging in complexity from a 36×36 m uniform square grid of 5 mCi Cu-64 sources to a configuration where regions of higher and zero activity are superimposed on a uniform baseline. We then present several example calculations for various detector systems, altitudes, array source spacings, and source patterns, and examine under what parameters it is possible to design a surrogate array source that is nearly indistinguishable from a truly continuous distributed source. In Part II, we will detail the design, manufacture, and testing of Cu-64 sealed sources at the Washington State University research reactor, discuss their deployment during the aerial measurement campaign, and present results from several measurements.

In this second part of a multi-paper series, we present results from outdoor aerial measurements of surrogate distributed gamma-ray sources. We detail the design, manufacture, and testing of 300 individual ~7 mCi Cu-64 sealed sources at the Washington State University research reactor, and their deployment in various source patterns (each comprising up to 100 point sources) during the aerial measurement campaign. We show the results of two such measurements, in which approximate source shapes and qualitative source intensities can be seen from the count rate vs. position plots, even without performing reconstructions. We also detail our efforts in ground-truthing the deployed sources and comparing measured gamma ray data to model predictions. In particular we compare measured vs. expected count data using the Poisson deviance formalism of Part I to evaluate whether the fielded surrogate point-source arrays “look like” their truly continuous distributed source analogues. More generally, we find that the point-source array technique provides high source placement accuracy, relative ease of quantifying the true source configuration, scalability to source dimensions of ≲100 m, ease of reconfiguration and removal, and relatively low dose to personnel. Finally, we consider potential improvements and generalizations of the point-source array technique for future measurement campaigns.

The addition of contextual sensors to mobile radiation sensors provides valuable information about radiological source encounters that can assist in adjudication of alarms. This study explores how computer-vision based object detection and tracking analyses can be used to augment radiological data from a mobile detector system. We study how contextual information (streaming video and LiDAR) can be used to associate dynamic pedestrians or vehicles with radiological alarms to enhance both situational awareness and detection sensitivity. Possible source encounters were staged in a mock urban environment where participants included pedestrians and vehicles moving in the vicinity of an intersection. Data was collected with a vehicle equipped with 6 NaI(Tl) 2 inch times 4 inch times 16 inch detectors in a hexagonal arrangement and multiple cameras, LiDARs, and an IMU. Physics-based models that describe the expected count rates from tracked objects are used to correlate vehicle and/or pedestrian trajectories to measured count-rate data through the use of Poisson maximum likelihood estimation and to discern between source-carrying and non-source-carrying objects. In this work, we demonstrate the capabilities of our source-object attribution approach as applied to a mobile detection system in the presence of moving sources to improve both detection sensitivity and situational awareness in a mock urban environment.

Static gamma-ray detector systems that are deployed outdoors for radiological monitoring purposes experience time- and spatially-varying natural backgrounds and encounters with man-made nuisance sources. In order to be sensitive to illicit sources, such systems must be able to distinguish those sources from benign variations due to, e.g., weather and human activity. In addition to fluctuations due to non-threats, each detector has its own response and energy resolution, so providing a large network of detectors with predetermined background and source templates can be an onerous task. Instead, we propose that static detectors use simple physics-informed algorithms to automatically learn the background and nuisance source signatures, which can them be used to bootstrap and feed into more complex algorithms. Specifically, we show that non-negative matrix factorization (NMF) can be used to distinguish static background from the effects of increased concentrations of radon progeny due to rainfall. We also show that a simple process of using multiple gross count rate filters can be used in real time to classify or “triage” spectra according to whether they belong to static, rain, or anomalous categories for processing with other algorithms. If a rain sensor is available, we propose a method to incorporate that signal as well. Two clustering methods for anomalous spectra are proposed, one using Kullback-Leibler divergence and the other using regularized NMF, with the goal of finding clusters of similar spectral anomalies that can be used to build anomaly templates. Finally we describe the issues involved in the implementation of some of these algorithms on deployed sensor nodes, including the need to monitor the background models for long-term drifting due to physical changes in the environment or changes in detector performance.

As more complex predictive models are used for gamma-ray spectral analysis, methods are needed to probe and understand their predictions and behavior. Recent work has begun to bring the latest techniques from the field of Explainable Artificial Intelligence (XAI) into the applications of gamma-ray spectroscopy, including the introduction of gradient-based methods like saliency mapping and Gradient-weighted Class Activation Mapping (Grad-CAM), and black box methods like Local Interpretable Model-agnostic Explanations (LIME) and SHapley Additive exPlanations (SHAP). In addition, new sources of synthetic radiological data are becoming available, and these new data sets present opportunities to train models using more data than ever before. In this work, we use a neural network model trained on synthetic NaI(Tl) urban search data to compare some of these explanation methods and identify modifications that need to be applied to adapt the methods to gamma-ray spectral data. We find that the black box methods LIME and SHAP are especially accurate in their results, and recommend SHAP since it requires little hyperparameter tuning. We also propose and demonstrate a technique for generating counterfactual explanations using orthogonal projections of LIME and SHAP explanations.

The ability to formulate maps of minimum detectable activities (MDAs) that describe the sensitivity of an ad hoc measurement that used one or more freely moving radiation detector systems would be significantly beneficial for the conduct and understanding of many radiological search activities. In a real-time scenario with a free-moving detector system, an MDA map can provide useful feedback to the operator about which areas have not been searched as thoroughly as others, thereby allowing the operator to prioritize future actions. Similarly, such a calculation could be used to inform subsequent navigation decisions of autonomous platforms. Here we describe a near real-time MDA mapping approach that can be applied when searching for point sources using detected events in a spectral region of interest while assuming a constant, unknown background rate. We show the application of this MDA mapping method to a real scenario, a survey of the interior of a small building using a handheld detector system. Repeated measurements with no sources and with ¹³⁷ Cs sources of different strengths yield results consistent with the estimated thresholds and MDA values; namely, that for background-only measurements no sources are seen above threshold anywhere in the scene, while when sources are present they are detected above the thresholds calculated for their locations.