Congress:
EuroSafe Imaging 2020
Keywords:
Multicentre study, Observational, Retrospective, Quality assurance, Diagnostic procedure, CT, Conventional radiography, Radioprotection / Radiation dose, Action 4 - Dose management systems
Authors:
T. Vanaudenhove, S. Taylor, A. Van Muylem, P. A. Gevenois, D. Tack
DOI:
10.26044/esi2020/ESI-05665
Description of activity and work performed
Local variabilities of DLP from two CT devices and DAP from three radiology tables were investigated using a bootstrapping method. Variabilities of national DLP-DRLs and DAP-DRLs were investigated according to the number of devices and sample size per device. Local and national variabilities were computed for common examinations (e.g. trunk, brain, lumbar spine) to ensure sufficient data. Variabilities for other procedures have to be considered but are expected to be higher than observed. No factors influencing the inherent variabilities (patient size, scanner type, diagnostic quality) were investigated.
Variabilities of local metrics depended on the sample size for each device. At the local level, for samples ≤ 20 data per device, variabilities of DLP were as high as 60% [1] of the median value and those of DAP as high as 175% [2]. From 300 to 600 CT data were needed to reach a variability of less than 10%. These numbers were reduced, between 150 and 250, if weight-selection (adults around 70 kg) was performed on the sample. For conventional radiology, between 1,000 and 2,000 data were needed to reach a 10% variability.
The number of data per device had no impact on the variabilities of DLP-DRLs and DAP-DRLs, provided that the number of data per device was higher than 20. These variabilities depended mainly on the number of devices. For CT scanners, the variabilities of DLP-DRLs ranged from 20% to 40% for a number of devices between 50 and 70 but was stabilized at a level of 10–20% if the number of devices was higher than 200 [3]. For conventional radiology, 200 devices were needed to reduce the variabilities of DAP-DRLs to less than 50% and 400 devices were needed to reduce them to 25%. A 10% variability could only be reached with thousands of devices [4].