Unfolding-assisted Bayesian quantification of radionuclide mixtures in NaI(Tl) spectra with overlapping gamma-ray...

Gamma-ray spectrometry using NaI(Tl) detectors is well suited for rapid radionuclide screening; however, the limited energy resolution of these detectors causes overlap among gamma-ray peaks emitted by radionuclide mixtures. This study proposes a unfolding-assisted Bayesian inference methodology to address the resulting challenges in nuclide identification and activity quantification in low-resolution gamma-ray spectra. A detector response matrix was constructed using Monte Carlo simulations followed by Gaussian energy broadening. The response matrix was used to unfold the observed spectrum, identify latent gamma-ray peaks, and generate a candidate nuclide set. The candidate nuclides were then incorporated into Bayesian inference to estimate activity posteriors and derive detection decisions. The framework was evaluated using synthetic radionuclide-mixture scenarios designed to represent radiological emergency conditions, where overlapping gamma-ray peaks complicate the quantification of accident-related radionuclides. Across all tested scenarios, the posterior mean estimates of accident-related radionuclide activities showed a maximum absolute relative bias below 9%, and the posterior decision analysis achieved detection sensitivities ranging from 75% to 100%. The results demonstrate that the proposed method can provide reliable activity quantification using low-resolution NaI(Tl) gamma-ray spectrometry in the presence of substantial peak overlap in spectra. Its automated workflow, from latent peak search to posterior activity decision, has the potential to support rapid radiological emergency response by reducing reliance on expert-driven spectral interpretation.