Effect of spectrum processing procedure on the linearity of EPR dose reconstruction in tooth enamel

Electron Paramagnetic Resonance (EPR) spectroscopy with tooth enamel is a widely used method of dosimetry. The accuracy of EPR tooth dosimetry depends on the spectrum processing procedure, the quality of which, in its turn, relies on instrumental noise and the signals from impurities. This is especially important in low-dose evaluation. The current paper suggests a method to estimate the accuracy of a specific spectrum processing procedure. The method is based on reconstruction of the radiation-induced signal (RIS) from a simulated spectrum with known RIS intensity. The Monte Carlo method was used for the simulations. The model of impurity and noise signals represents a composite residual spectrum (CRS) obtained by subtraction of the reconstructed RIS and the native background signal (BGS) from enamel spectra measured in HMGU (Neuherberg, Germany) and IMP (Yekaterinburg, Russia). The simulated spectra were deconvoluted using a standard procedure. The method provides an opportunity to compare the simulated “true” RIS with reconstructed values. Two modifications of the EPR method were considered: namely, with and without the use of the reference Mn²⁺ signals. It was observed that the spectrum processing procedure induces a nonlinear dose response of the reconstructed EPR amplitude when the height of the true RIS is comparable with the amplitudes of noise-like random splashes of CRS. The area of nonlinearity is below the limit of detection (DL). The use of reference Mn²⁺ signals can reduce the range of nonlinearity. However, the impact of the intensities of CRS random signals on nonlinearity is two times higher than the one observed when the reference signals were not used. The reproducibility of the software response is also dependent on both the amplitude of the CRS and the use of a reference signal, and it is also two times more sensitive to the amplitude of the CRS. In most EPR studies, all of the data are used, even those for which the dose value is lower than the DL. This study shows that low doses evaluated with the help of linear dose–response can be significantly overestimated. It is recommended that linear dose response calibration curves be constructed using only data above the DL. Data below the DL should be interpreted cautiously.