Shapley effects to identify the most influential input parameters of organ dose estimation following CT scans in childhood.

Abstract

A current methodological challenge in radiation epidemiology is to account for dose uncertainty in risk estimation. An important preliminary step can be to identify the most influential input parameters involved in dose estimation, before modelling and accounting for their related uncertainty in radiation-induced health risk estimates. In this work, a variance-based global sensitivity analysis (GSA) was performed with Shapley effects and, for the first time in radiation dosimetry, to rank by influence the input parameters of a software conventionally used to estimate organ doses arising from computed tomography (CT) scan exposure: NCICT (National Cancer Institute Dosimetry System for CT) (version 1.0). We focused on brain and red bone marrow (RBM) dose estimation in the French CT cohort and our GSA was performed for three types of examination, classified according to the scanned body region: head, chest, and abdominopelvic. Among the eight uncertain input parameters involved in dose estimation, different influential parameters were highlighted depending on both the scan type and the considered organ dose: these are consistent with the underlying dose calculation methodology. For head CT examinations, technical parameters used for image acquisition including x-ray tube current-time product (mAs) and tube potential (kVp), as well as the end of the scanned body area contributed to about 70% of the variance of the brain dose estimate, while mAs, kVp, the start and end of the scanned body area, and age influenced the RBM dose. The brain and RBM doses delivered during chest CT examinations were mainly influenced by mAs, pitch, and kVp. Finally, kVp was the only input parameter with a significant influence on both the RBM and the brain dose for abdominopelvic CT examinations: it explained more than 30% of the variance of these organ doses. From a clinical perspective, our findings may support optimisation of paediatric CT protocols.