TY - JOUR
T1 - Experimental tests of Gamma-ray Localization Aided with Machine-learning (GLAM) capabilities
AU - Durbin, Matthew
AU - Sheatsley, Ryan
AU - McDaniel, Patrick
AU - Lintereur, Azaree
N1 - Publisher Copyright:
© 2022 Elsevier B.V.
PY - 2022/9/1
Y1 - 2022/9/1
N2 - Gamma-ray Localization Aided with Machine-learning (GLAM) utilizes an array of four rectangular NaI(Tl) detectors, and data processed with a k-nearest neighbor machine learning model, to predict the location of gamma-ray sources in stationary scenarios. This work demonstrates GLAM capabilities to predict real source locations when trained purely on simulated data. To account for efficiency differences not captured in simulations, a simple calibration methodology that improves agreement between simulated and experimental data on multi-detector arrays is also introduced. The GLAM approach was tested using two different sets of input features. The first set of input features consisted of the total counts in each detector. The second set of input features consisted of the counts from each detector in the photopeak and the Compton continuum, which resulted in better performance than simply using the total counts. For one-minute measurements of a 40 μCi 137Cs source up to 500 cm away, the mean absolute errors (MAEs) for the angular and radial predictions were 3.5°and 10.0%, respectively. For one-minute measurement times of a 6 μCi 60Co source up to 250 cm away, the MAEs for the angular and radial predictions were 2.8°and 12.7%, respectively.
AB - Gamma-ray Localization Aided with Machine-learning (GLAM) utilizes an array of four rectangular NaI(Tl) detectors, and data processed with a k-nearest neighbor machine learning model, to predict the location of gamma-ray sources in stationary scenarios. This work demonstrates GLAM capabilities to predict real source locations when trained purely on simulated data. To account for efficiency differences not captured in simulations, a simple calibration methodology that improves agreement between simulated and experimental data on multi-detector arrays is also introduced. The GLAM approach was tested using two different sets of input features. The first set of input features consisted of the total counts in each detector. The second set of input features consisted of the counts from each detector in the photopeak and the Compton continuum, which resulted in better performance than simply using the total counts. For one-minute measurements of a 40 μCi 137Cs source up to 500 cm away, the mean absolute errors (MAEs) for the angular and radial predictions were 3.5°and 10.0%, respectively. For one-minute measurement times of a 6 μCi 60Co source up to 250 cm away, the MAEs for the angular and radial predictions were 2.8°and 12.7%, respectively.
UR - http://www.scopus.com/inward/record.url?scp=85131748413&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85131748413&partnerID=8YFLogxK
U2 - 10.1016/j.nima.2022.166905
DO - 10.1016/j.nima.2022.166905
M3 - Article
AN - SCOPUS:85131748413
SN - 0168-9002
VL - 1038
JO - Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment
JF - Nuclear Instruments and Methods in Physics Research, Section A: Accelerators, Spectrometers, Detectors and Associated Equipment
M1 - 166905
ER -