Automated ejection fraction determination from gated myocardial FDG-PET data

Abstract

Background. The aim of this study was to determine the potential of the automated calculation of the left ventricular ejection fraction from gated myocardial positron emission tomography (PET) scans.

Methods. We retrospectively analyzed the data of 20 patients who underwent both gated fluorine 18 deoxyglucose (FDG)-PET and equilibrium radionuclide angiography (ERNA). Gated PET data were analyzed by 2 independent programs (ie, quantitative gated single photon emission computed tomography [QGS]) originally developed for gated single photon emission computed tomography studies and functional polarmap (FPM) originally developed for the analysis of (functional) dynamic PET studies. ERNA data were used as the gold standard.

Results. Both QGS and FPM left ventricular ejection fraction results correlated highly with ERNA (y = 0.90 × x−5.9, r = 0.86, P < .0001; y = 0.80 × x+3.3, r = 0.84, P < .0001, respectively). The correlation between FPM and QGS left ventricular ejection fraction results was even higher (y = 0.89 × x+8.6, r = 0.97, P < .0001). Bland-Altman plots showed systematic differences in the left ventricular ejection fraction of −9.6% ± 7.5% (QGS vs ERNA), −3.8% ± 7.8% (FPM vs ERNA), and −5.8% ± 3.5% (QGS vs FPM). Further comparison of the left ventricular volumes revealed systematic difference between QGS and FPM. Our results indicate that the correlation between the different left ventricular ejection fractions shows little sensitivity to errors in the left ventricular volumes; however, the exact relationship is influenced by these errors.

Conclusion. It is concluded that the automated determination of the left ventricular ejection fraction from gated PET data has significant potential; its results are highly and significantly correlated with ERNA. However, the methods presented here require additional calibration before final accuracy and clinical applicability can be determined.