Development of an automated isotope identification algorithm based on second derivative and Bayesian statistics methods using medium energy resolution scintillation detectors

Radioisotope identification devices(RID) play a crucial role in detection and identification of illicit trafficking of radioactive/nuclear materials in nuclear security and nuclear safeguards. These devices utilize various algorithms for automated isotope identification(ID) without the need for expert intervention. In this study, an automated algorithm for real-time isotope identification is presented. The algorithm employs a second-derivative-based peak detection and a Bayesian-statistics-peak based ID approach. To demonstrate the suitability of the developed algorithm, it was applied to the gamma-ray spectra acquired with a medium energy -resolution LaBr 3 (Ce) detector. In addition to point sources 60 Co, 109 Cd, 22 Na, 137 Cs, 241 Am, 152 Eu, and 133 Ba, the algorithm was also tested on the more complex gamma-ray spectra obtained from low enriched uranium reference materials 171 (EC-NRM171), and natural uranium and thorium minerals such as BL-2, BL-3, BL-4A, BL-5, RGU and OKA-2. To evaluate the performance of the algorithm, total scoring (ST) is calculated. For identification of 22 Na, 60 Co, and 137 Cs isotopes, the posterior probabilities were estimated to be greater than 99 %. For 133 Ba, 152 Eu, and 241 Am, the isotopes they were also correctly identified with higher posterior probabilities ranged from 92 % to 95 %. The developed algorithm successfully identified the isotopes contained in U-Th ore samples with a 100 % total score. Additionally, the performance evaluation of the results obtained with Certified Reference Uranium Materials also demonstrated 100 % score. For automatic ID, the photopeak-based Bayesian method, combined with the Mariscotti's peak detection method has great potential for real-time ID when implemented in RID devices.