Free-moving Quantitative Gamma-ray Imaging

Abstract

The ability to map and estimate the activity of radiological source distributions in unknown three-dimensional environments has applications in the prevention and response to radiological accidents or threats as well as the enforcement and verification of international nuclear non-proliferation agreements. Such a capability requires well-characterized detector response functions, accurate time-dependent detector position and orientation data, a digitized representation of the surrounding 3D environment, and appropriate image reconstruction and uncertainty quantification methods. We have previously demonstrated 3D mapping of gamma-ray emitters with free-moving detector systems on a relative intensity scale using a technique called Scene Data Fusion (SDF). Here we characterize the detector response of a multi-element gamma-ray imaging system using experimentally benchmarked Monte Carlo simulations and perform 3D mapping on an absolute intensity scale. We present experimental reconstruction results from hand-carried and airborne measurements with point-like and distributed sources in known configurations, demonstrating quantitative SDF in complex 3D environments.

Design and deployment of radiological point-source arrays for the emulation of continuous distributed sources

“…7) detector response validation: we showed good agreement between data and our response models for the MiniPRISM detector in Ref. [22]. We performed an additional validation study with NG-LAMP and Na-22 check sources (which also emit at 511 keV), and found agreement between expected and measured effective areas to within ∼10%.…”

Improved Gamma-Ray Point Source Quantification in Three Dimensions by Modeling Attenuation in the Scene

“…Although the data are individually read out by detector, for this analysis the detector events were collected together as if the system were a single, monolithic detector to maintain simplicity and keep computation times manageable while demonstrating efficacy in modeling attenuation. A measurement-validated simulated response function in all 4π steradians was used when performing PSL [21], and the simulations used the actual configuration of 58 detectors used in the measurement. Only the responses for the 54 detectors that had adequate performance at the time were summed together and used in the analysis.…”

“…However, the processing here has been done on a single CPU, which is the worst case scenario. In principle, the ray casting can be parallelized and run on GPUs, and a single PSL optimization has already been implemented on a GPU [21]. A better optimization procedure than a brute force grid search and better initial guesses based on contextual information should also be able to speed up evaluation time, and constraints from computer vision classification or operator input could further reduce the computational complexity and afford nearly real-time inference.…”

“…Because of the computational complexity and the difficulty of obtaining a representative model of the measurement geometry and incorporating it into an algorithm, attenuation by intervening material is seldom considered [15], [16]. Only recently have source localization methods been applied in 3D without attenuation [18], [19], and our recent work has demonstrated accurate quantitative reconstruction in 3D with freely moving systems in cases when attenuation can be ignored [14], [20], [21]. It is only when the full geometrical complexity of source localization problems is considered that a general attenuation model becomes a logical next step.…”

“…The solution framework we will use is referred to as Point Source Likelihood (PSL) [14], [20], [21]. PSL requires the knowledge of a detector's position and orientation in 3D as it collects gamma-ray events, as well as knowledge of the detector's angular response.…”