An inversion-forward method for γ-ray spectrum background subtraction based on the ML-EM inversion algorithm

γ-ray spectrum analysis is one of the widely used techniques in the field of nuclear radiation detection. Background subtraction from γ-ray spectra can improve the quantitative accuracy of γ-emitting radionuclides. To address the limitation of traditional methods in accurately subtracting backgrounds from complex γ-ray spectra, particularly those containing overlapping peaks, this paper proposes an inversion-forward method for γ-ray spectrum background subtraction based on the Maximum Likelihood-Expectation Maximization (ML-EM) inversion algorithm. A 4″ × 4″×16″ thallium-doped sodium iodide (NaI(Tl)) detector was used to acquire γ-ray spectra containing information from ²⁴¹Am, ¹³⁷Cs, ⁶⁰Co, ¹⁵²Eu, ¹³³Ba, and ²²Na. Focusing on ²⁴¹Am, ¹³⁷Cs, and ⁶⁰Co while treating the other nuclides as interference sources, the proposed method was applied for background subtraction. The results were compared with those obtained using the Maestro32 software and the Statistics-sensitive Nonlinear Iterative Peak-clipping (SNIP) method. The results demonstrate that the background subtraction accuracy of the proposed method is significantly superior to the other two methods. After background subtraction, the maximum absolute deviation of the calculated activities of ²⁴¹Am, ¹³⁷Cs, and ⁶⁰Co (where the ²⁴¹Am characteristic peak overlaps with that of ¹³³Ba, and the ⁶⁰Co characteristic peaks overlap with those of ¹⁵²Eu, ¹³³Ba, and ²²Na) from the reference values is only 5.6%, compared to 48.8% for Maestro32 and 72.5% for the SNIP method.