Background correction in first-pass radionuclide angiography: comparison of several approaches

This study was designed to test the comparative accuracy of several commonly used background correction techniques in first-pass radionuclide angiography (FPRNA). Thirty patients underwent FPRNA and single plane contrast angiography (CA) within 1 hr of each other. The left ventricular ejection fractions (LVEF) calculated from the different background subtraction approaches to FPRNA were compared to the CA LVEF. When applied to a representative cycle, a horseshoe-shaped background region of interest (BKROI) underestimated LVEF (p less than 0.005, r = 0.91, s.e.e. = 0.06) while a ring shaped BKROI adjusted at end-systole for aortic valve motion insignificantly overestimated LVEF (p = NS, r = 0.91, s.e.e. = 0.07). A lung background approach applied to a representative cycle gave the best correlation with CA (p = NS, r = 0.96, s.e.e. = 0.04). Without using a representative cycle, time-activity curves from a horseshoe-shaped BKROI and the LV ROI were created and the LV curve was normalized to the peak counts in the BKROI curve. LVEF calculated from the normalized curve correlated favorably with CA LVEF (p = NS, r = 0.91, s.e.e. = 0.08). The influence of some recently described improvements in representative cycle generation are also documented.