Automated size-specific dosimetry for chest posterior–anterior projection radiography

Abstract

IntroductionChest radiography is a frequently performed examination, and therefore, accurate patient dosimetry is important. One means of increasing dosimetric accuracy is through the use of size-specific dosimetry, and the aim of this work was to determine which patients would benefit from such a methodology applied in chest posterior–anterior (PA) projection radiography.MethodsA database of 44 voxel models was created from CT scans, representing adults with BMI from underweight, normal, overweight, and obese classes. Monte Carlo simulation was performed to generate dose conversion factors for each model. Correlation curves of the dose conversion factors with water equivalent diameter (WED) of the 44 voxel phantoms were obtained. A total of 8,536 chest PA examinations acquired between 2015 and 2019 using three X-ray systems were then retrospectively collected; 4,748 men (median age: 61 ± 21 years) and 3,788 women (median age: 60 ± 23 years). The WED of the patients was estimated from the ratio of detector air-kerma to incident air-kerma, using parameters in the DICOM header of the images. For all patients, a size-specific conversion factor was selected automatically using the relation between WED and dose conversion factor determined for the phantoms. The size-specific organ doses and effective doses were calculated based on these conversion factors, and then compared to the standard effective dose calculation with a paired test.ResultsThe lung doses ranged from 7 μGy to 96 μGy and had a good correlation with patient size in terms of WED (p < 0.01, R² between 0.52 and 0.77). The doses for thyroid ranged from 2 μGy to 42 μGy and correlated strongly with the patient size (p < 0.01, R² between 0.65 and 0.85). Breast doses ranged from 2 μGy to 13 μGy, and the correlation with the patient size was weak (p < 0.01, R2 between 0.01 and 0.28). The size-specific effective dose ranged from 4 μSv to 42 μSv. The difference between the size-specific and standard effective dose ranged from −35% to 69% (p < 0.01), with differences exceeding ± 20% for 37% of the cases.DiscussionIn conclusion, the method presented in this study enables automated size-specific dosimetry, within the proposed maximum deviation of ± 20%, and should be considered for routine application.