Accuracy of Specification of Dose Reference Values and Detail Detectability by a Statistical Phantom

The introduction of modern electronic image sensors and image processing techniques has tightened the link between dose and image quality. A properly exposed image is made with any dose, even at the cost of image quality. Therefore a test is needed to maintain minimum 'reference' image quality when determining reference doses. The underlying theory of image quality and detail detectability is outlined. Reference image quality can be expressed in terms of a minimum signal-to-noise ratio of a given detail. The accuracy of detail detectability testing with the statistical phantom for a storage phosphor system has been investigated. For four observers, each with 50 answers, ROC curves were calculated by the ROCFIT program. A simplified score-sum method of evaluation recommended for the ALVIM statistical phantom is then employed and the accuracy of both methods is evaluated. It can be shown that detail detectability tests using a statistical phantom are suitable for the needs of quality assurance from constancy to acceptance testing. The accuracy of ROC tests would be sufficient to specify sensitivity values for modern imaging systems by measuring the dose required for visualising a detail. (author)