We aim to improve the accuracy of electron paramagnetic resonance (EPR)-based in vivo tooth dosimetry using the relationship between tooth geometry and radiation-induced signals (RIS). A homebuilt EPR spectrometer at L-band frequency of 1.15 GHz originally designed for non-invasive and in vivo measurements of intact teeth was used to measure the RIS of extracted human teeth. Twenty human central incisors were scanned by microCT and irradiated by 220 kVp x-rays. The RISs of the samples were measured by the EPR spectrometer as well as simulated by using the finite element analysis of the electromagnetic field. A linear relationship between simulated RISs and tooth geometric dimensions, such as enamel area, enamel volume, and labial enamel volume, was confirmed. The dose sensitivity was quantified as a slope of the calibration curve (i.e., RIS vs. dose) for each tooth sample. The linear regression of these dose sensitivities was established for each of three tooth geometric dimensions. Based on these findings, a method for the geometry correction was developed by use of expected dose sensitivity of a certain tooth for one of the tooth geometric dimensions. Using upper incisors, the mean absolute deviation (MAD) without correction was 1.48 Gy from an estimated dose of 10 Gy; however, the MAD corrected by enamel area, volume, and labial volume was reduced to 1.04 Gy, 0.77 Gy, and 0.83 Gy, respectively. In general, the method corrected by enamel volume showed the best accuracy in this study. This homebuilt EPR spectrometer for the purpose of non-invasive and in vivo tooth dosimetry was successfully tested for achieving measurements in situ. We demonstrated that the developed correction method could reduce dosimetric uncertainties resulting from the variations in tooth geometric dimensions.

We aim to develop a dose assessment method compensating for quality factors (Q factor) observed during in vivo EPR tooth dosimetry. A pseudo-in-vivo phantom made of tissue-equivalent material was equipped with one each of four extracted human central incisors. A range of Q factors was measured at tooth-depths of −2, 0, and 2 mm in the pseudo-in-vivo phantom. In addition, in vivo Q factors were measured from nine human volunteers. For the dose-response data, the above four sample teeth were irradiated at 0, 1, 2, 5, and 10 Gy, and the radiation-induced signals were measured at the same tooth-depths using an in vivo EPR tooth dosimetry system. To validate the method, the signals of two post-radiotherapy patients and three unirradiated volunteers were measured using the same system. The interquartile range of the Q factors measured in the pseudo-in-vivo phantom covered that observed from the human volunteers, which implied that the phantom represented the Q factor distribution of in vivo conditions. The dosimetric sensitivities and background signals were decreased as increasing the tooth-depth in the phantom due to the decrease in Q factors. By compensating for Q factors, the diverged dose-response data due to various Q factors were converged to improve the dosimetric accuracy in terms of the standard error of inverse prediction (SEIP). The Q factors of patient 1 and patient 2 were 98 and 64, respectively, while the three volunteers were 100, 92, and 99. The assessed doses of patient 1 and patient 2 were 2.73 and 12.53 Gy, respectively, while expecting 4.43 and 13.29 Gy, respectively. The assessed doses of the unirradiated volunteers were 0.53, 0.50, and − 0.22 Gy. We demonstrated that the suggested Q factor compensation could mitigate the uncertainty induced by the variation of Q factors.

No abstract