Achieving High Resolution Gamma-Ray Spectra Using Spectrum Data Obtained from NaI(Tl) Detector by Multi-Output Regressor-Chain Structure Based on SVR

This paper proposes a novel approach for generating high-resolution energy spectra using cost-effective Sodium iodide Thallium activated (NaI(Tl)) detectors. It employ a multi-output regression chain structure based on support vector regression (SVR) to map NaI(Tl) spectra to their corresponding HPGe spectra. The suggested framework utilizes a regression chain strategy to enhance regression models that lack support for multi-output regression. This involves initially employing one regressor for each energy channel of the HPGe spectrum. Subsequently, multiple regressors are integrated to forecast all energy channels of HPGe spectrum. Each regressor in the chain receives the entire NaI spectrum as input. Then, for each subsequent regressor, input is further augmented by concatenating the outputs of all preceding regressors in the chain. Despite being trained on a limited radioisotope library, the model exhibits exceptional performance across diverse measured test spectra containing multiple radioisotopes. Among the various kernel functions employed (linear, radial basis function (RBF), and polynomial), the RBF and polynomial kernels yielded superior performance compared to the linear kernel. By enabling HPGe spectrum prediction using NaI(Tl) detectors, this study highlights a significant advancement in radiation detection capabilities, addressing cost and operational considerations.