Characterization of Radiation Risk across a Clinical CT Patient Population: Comparison across 12 Risks Metrics
PURPOSE Ascertaining radiological procedure radiation burden is essential for justification, optimization, and personalization of the procedure. While the exact radiation risk for an individual exam is unknowable, various risk-related figures have been used as surrogates, e.g., device output metrics such as CTDI, DLP, SSDE, and hypothetical constructs such as Effective Dose (ED) and Risk Index (RI) that take into account specific organ risks, age, and gender factors. Purpose of this study was to compare how twelve different radiation risk metrics characterize the radiation burden across a set of clinical CT examinations differently. METHOD AND MATERIALS This IRB-approved study included 265 abdominopelvic exams with contrast. Organ doses were estimated using Monte Carlo methods. The following risk metrics were calculated using previously validated methods: CTDIvol, DLP, SSDE, DLP-based ED (EDk), organ-dose-based ED (EDOD), dose to a defining organ (ODD to stomach), organ-dose-based RI (RI), and RI for a reference 20 y.o. patient (RIr). Additional metrics of ODD,0, ED0, and RI0 were calculated for a reference patient (ICRP 110). Lastly, inspired by the ICRP, an adjusted ED (ED') was computed as the product of RI/RIr and EDOD. A linear regression was applied to each metric dependency to the patient water equivalent diameter (WED). Fit slopes (FS) and relative interquartile ranges (rIQR) at 30 cm reference WED were calculated and normalized to those of RI, which was assumed to be the actual patient risk closest surrogate. RESULTS Results showed significant differences between the metrics. EDOD exhibited closest concordance with RI, followed by ODD. Normalized FS ranged between 0.99 (ED') to 2.24 (DLP), and normalized rIQR ranged between 0.30 (DLP) to 4.03 (ED'). CONCLUSION Different risk metrics lead to different characterization of population risk, particularly risk overestimation for large patients and underestimation for small patients. Furthermore, the risk variability across a population is underestimated for most variables except ED', which exhibits more variability and radiation risk individualization. Care should be exercised in drawing risk predictions from unrepresentative risk metrics applied to a population. CLINICAL RELEVANCE/APPLICATION Different risk metrics can lead to different risk predictions. There is a need to standardize risk metrology for proper justification and optimization of radiological procedures.
