Assessment of scintigram quality

Abstract

The application of standard psychophysical techniques to the measurement of the quality of noisy images is shown using examples taken from clinical radionuclide imaging. In clinical radionuclide imaging the distribution of a radiopharmaceutical injected into a patient is mapped in vivo by using a scintillation detector to record the emitted gamma rays. Areas of above or below average radiopharmaceutical concentration may indicate clinical abnormalities, such as tumours. As the images have only a few hundred photons/cm2, random (Poisson) fluctuations in intensity are large and may mask changes in radiopharmaceutical concentration. The quality of these images is assessed by using either the method of constant stimulus or signal-detection theory to measure an observer's ability to detect small changes in image intensity. The method of constant stimulus is used to compare analogue with digital images. It is shown that smaller changes in intensity can be detected on digital images than on the analogue ones. Signal detection theory is used to measure the effect of filtering on image quality. For a given rate of false positive responses, the filtered images produce a higher true positive response rate than for the unfiltered ones.