Echocardiographic validation of gated SPECT ventricular function measurements.

UNLABELLED

Left ventricular (LV) volumes are valuable prognostic indicators in the management of coronary artery disease and traditionally have been obtained by x-ray contrast angiography or echocardiography. There now are several scintigraphic methods to compute volumes that are based on different LV modeling assumptions. Both the reasons that calculations from different nuclear techniques can disagree with one another and the relationship of these values to the more conventional echocardiographic measurements must be investigated thoroughly for calculations to be interpretable for individual patients.

METHODS

Echocardiographic volumes were determined in 33 retrospective subjects with coronary artery disease (mean age, 61 +/- 12 y; 42% men; 70% with abnormal perfusion and 58% with abnormal segmental wall motion) using the modified Simpson's rule technique applied to digitized apical 4-chamber and apical 2-chamber views of 4 averaged heartbeats. These volumes were compared with those from 3 gated SPECT methods based on Simpson's rule LV modeling similar to standard echocardiographic algorithms (SPECT EF from St. Luke's-Roosevelt Hospital) (method 1), Gaussian myocardial count profile curve fitting (QGS from Cedars-Sinai Medical Center) (method 2), and an endocardial model based on perfusion sampling and count-based thickening (Cardiac Toolbox from Emory University) (method 3).

RESULTS

By ANOVA, there were no significant differences among ejection fractions (EFs), but there were for volumes. Paired t test analysis showed volumes from methods 2 and 3 to be significantly larger than echocardiographic volumes and larger than those of method 1. Linear regression analysis comparing gated SPECT and echocardiographic volumes showed a nearly identical strong correlation (r = 0.92; P < 0.000001) for all 3 methods. Excellent correlation also was found among gated SPECT volumes from the 3 methods (r = 0.94). Bland-Altman analysis and t tests showed that method 1 volumes (70 +/- 61 mL) were the same as for echocardiography (77 +/- 55 mL), but volumes were overestimated by method 2 (105 +/- 74 mL) and method 3 (127 +/- 92 mL), particularly for larger volumes. Pearson coefficients for EFs compared with echocardiography were r = 0.82, 0.75, and 0.72 for methods 1-3, respectively. EFs correlated strongly among the 3 gated SPECT methods (r = 0.86-0.92). The Fisher z test showed no differences among these methods for any of the volume or EF linear correlation analyses.

CONCLUSION

All gated SPECT parameters correlated well with echocardiographic values. However, the gated SPECT method for which underlying assumptions most closely resembled those commonly used in echocardiography produced mean volume values closest in agreement with echocardiographic measurements.