Fourier temporal interpolation improves electrocardiograph-gated myocardial perfusion SPECT

The optimal temporal sampling rate in electrocardiograph-gated myocardial SPECT is questionable: low rates, typically 8 frames per cardiac beat (8fr/cb), favor image quality, whereas high rates, typically 16 frames per cardiac beat (16fr/cb), favor the accuracy of left ventricular (LV) functional parameters. We examined whether Fourier temporal interpolation (FTI) from 8fr/cb to 16fr/cb can combine the advantages of low and high rates.

In 34 patients imaged after stress injection of (99m)Tc-sestamibi, 4 sets of reconstructed gated slices were compared: a raw 16fr/cb acquisition (R16), a raw 8fr/cb acquisition (R8), a 16fr/cb set obtained by FTI of 8fr/cb projections (IP), and a 16fr/cb set obtained by FTI of 8fr/cb reconstructed slices (IS). LV ejection fraction (LVEF), end-diastolic volume (EDV), and end-systolic volume (ESV) obtained from the final LV volume curve were compared for the 4 datasets. Deviation of the whole LV volume curve was quantified for IP and IS with respect to R16. Image quality was evaluated by consensus reading of end-diastolic slices of the 4 sets. For R16, IP, and IS, cine display fluidity was quantified by a roughness index calculated from the LV volume curve.

No differences in EDVs or ESVs were found among R16, IP, and IS, whereas R8 gave smaller EDVs and larger ESVs. LVEF was lower with R8, IP, and IS than with R16: -3.9%, -1.2%, and -1.3%, respectively. The LV volume curve was closer to R16 with IP than with IS. Image quality was better with IP and IS than with R8 and better with R8 than with R16. Cine display fluidity was better with IP than with R16 and better with R16 than with IS.

FTI improved image quality not only over that provided by R16 but even over that provided by R8. The sole worsened LV functional parameter was LVEF, which was slightly underestimated with respect to that estimated by R16. Of the 2 FTI variants, IP was superior to IS for cine display fluidity and accuracy of the LV volume curve with respect to the data obtained with R16. Therefore, FTI to 16fr/cb performed before reconstruction on a pixel-by-pixel basis on 8fr/cb projections improves image quality and cine display fluidity over those of both R8 and R16 acquisitions at the sole cost of a 1% underestimation of LVEF.