Model dependence of gated blood pool SPECT ventricular function measurements

Abstract

Background: Calculation differences between various gated blood pool (GBP) single photon emission computed tomography (SPECT) (GBPS) algorithms may arise as a result of different modeling assumptions. Little information has been available thus far regarding differences for right ventricular (RV) function calculations, for which GBPS may be uniquely well suited.

Methods and results: Measurements of QBS (Cedars-Sinai Medical Center, Los Angeles, Calif) and BP-SPECT (Columbia University, New York, NY) algorithms were evaluated. QBS and BP-SPECT left ventricular (LV) ejection fraction (EF) correlated strongly with conventional planar-GBP LVEF for 422 patients (r = 0.81 vs r = 0.83). QBS correlated significantly more strongly with BP-SPECT for LVEF than for RVEF (r = 0.80 vs r = 0.41). Both algorithms demonstrated significant gender differences for 31 normal subjects. BP-SPECT normal LVEF (67% +/- 9%) was significantly closer to values in the magnetic resonance imaging (MRI) literature (68% +/- 5%) than QBS (58% +/- 9%), but both algorithms underestimated normal RVEF (52% +/- 7% and 50% +/- 9%) compared with the MRI literature (64% +/- 9%). For 21 patients, QBS correlated similarly to MRI as BP-SPECT for LVEF (r = 0.80 vs r = 0.85) but RVEF correlation was significantly weaker (r = 0.47 vs r = 0.81). For 16 dynamic phantom simulations, QBS LVEF correlated similarly to BP-SPECT (r = 0.81 vs r = 0.91) but QBS RVEF correlation was significantly weaker (r = 0.62 vs r = 0.82). Volumes were lower by QBS than BP-SPECT for all data types.

Conclusions: Both algorithms produced LV parameters that correlated strongly with all forms of image data, but all QBS RV relationships were significantly different from BP-SPECT RV relationships. Differences between the two algorithms were attributed to differences in their underlying ventricular modeling assumptions.