Statistical analysis for obtaining optimum number of CT scanners in patient dose surveys for determining national diagnostic reference levels.

<bold>Objectives: </bold>To statistically determine an 'optimum number of CT scanners' for obtaining 'diagnostic reference levels' (DRLs) in CT examinations as close as possible to 'ideal DRLs' when all available CT scanners are considered.<bold>Methods: </bold>First, six 'ideal DRLs' (CTDIVol and DLP) were determined for head, chest and abdomen/pelvis examinations by using patient-dose survey data of 100 CT scanners of different models in Tehran. Then, a 'random sampling method' was applied to different percent fractions of patient dose data of 100 CT scanners. The percent differences (PD) of the DRLs obtained from 'ideal DRLs' and their coefficients of variation (CVs) were calculated. The 'optimum number of CT scanners' determined met those of 'ideal DRL' criteria; i.e. precision (CV ≤ 10%) and accuracy (PD ≤ 10%).<bold>Results: </bold>'Optimum number of CT scanners' for determining DRLs as close as possible to 'ideal DRLs', fulfilling the stated criteria, is 43 instead of using 100.<bold>Conclusion: </bold>'Optimum number of CT scanners' for obtaining DRLs as close as possible to 'ideal DRLs' was determined. This optimum number can be effectively applied in patient-dose survey situations with limited resources in a time- and cost-effective manner.<bold>Key Points: </bold>• Ideal DRLs were determined by a CT patient-dose survey applied to available scanners. • 'Optimum number of CT scanners' statistically determined for DRLs is 43%. • Optimum number can be used for DRLs as if 'ideal DRLs' were determined by all scanners. [ABSTRACT FROM AUTHOR]