The occurrence of false-positive technetium-99m sestamibi bull's eye defects in different reference databases

Abstract

Myocardial perfusion single-photon emission tomographic (SPET) imaging has been shown to be sensitive in the detection of coronary artery disease (CAD), whereas its specificity has been suboptimal. The aim of the present study was to study the frequency of abnormal bull's eye perfusion defects in a large age-stratified group of healthy subjects undergoing myocardial SPET assessed by comparison with two existing commercially available reference databases. One hundred and twenty-eight healthy volunteers (76 males and 52 females) with a less than 5% likelihood of CAD underwent rest and exercise technetium-99m sestamibi SPET. The false-positive response rate, defined as a significant reversible defect, was 12% when compared to the CEqual database and 29% when compared to the Cedars-Sinai database. With the CEqual program, rest defects occurred in 12% of the subjects. Defects occurred more often in women than in men, but the difference did not attain statistical significance. Significant defects were most frequent in the inferior wall and in women in the anterior wall as well. The distribution of defects was independent of age. Our results suggest that the specificity of 99mTc-sestamibi myocardial SPET using commercially available reference files is suboptimal. The risk of obtaining a false-positive test result in subjects undergoing 99mTc-sestamibi myocardial SPET with a very low likelihood of CAD was higher than anticipated. With both reference files false-positive test results were most frequently observed in the inferior wall. Our data suggest that commercial reference files for myocardial SPET need to be optimised, and should be used with caution. The use of attenuation correction may prove to be a major step forward.

Quantitative digital evaluation of myocardial exercise thallium‐201 single‐photon emission tomography in post‐menopausal women

“…The differences in reference values can be partly explained by the fact that our system was corrected for attenuation in the reconstruction algorithm (Chang, 1978). Bouvier et al (1997) reported that linear or uniform correction for attenuation is not perfect, but it was used during this study and the optimal method is not yet established (Toft et al, 1997). However, this can only explain our higher values inferiorly immediately post exercise but not why there was no difference after delay.…”

“…Quantitative computerized digital analysis with bull's-eye processing theoretically provides an independent evaluation, but the camera, the operatordependent steps and the anatomy may introduce artefacts and spoil the result (Eisner et al, 1988;Beller, 1994;DePuey, 1994;Cohen et al, 1996;Bouvier et al, 1997;Toft et al, 1997). It has, however, been stated that quantitative scintigraphy improves the diagnostic accuracy in individual coronary arteries, particularly in patients with multivessel disease (Berger et al, 1981).…”

“…It has, however, been stated that quantitative scintigraphy improves the diagnostic accuracy in individual coronary arteries, particularly in patients with multivessel disease (Berger et al, 1981). The creation of a software phantom database based on clinical studies has been suggested for quality assurance of software used for analysis (Bourguignon et al, 1993) and the commercially available reference ®les are inadequate, particularly in women (Toft et al, 1997). Individual experience of a certain technique as well as available clinical information in¯uence the interpretation of a scintigram (Simons et al, 1994).…”

Diagnostic accuracy of quantitative cardiac MRI evaluation compared to stress single-photon-emission computed tomography

“…Inferior wall attenuation is a common problem in the interpretation of SPECT [27,28]. Previous studies have suggested that cMRI is more useful than SPECT in differentiating artifactual attenuation of the inferior wall [29].…”

Scandinavian test of artificial neural network for classification of myocardial perfusion images

“…In a number of neural network studies (Scott & Palmer, 1993;Tourassi et al, 1993;Fisher et al, 1996;Scott et al, 1996), the inputs to the networks were expert's visual pre-classi®cation of the images. This criterion is adopted from a study by Toft et al (1997a) and gives a relatively high speci®city similar to that of the experienced physician. In a neural network study on pulmonary ventilation±perfusion imaging, Scott et al (1996) showed how variability between different physicians affected the network performance.…”