Potential impact of hybrid CZT SPECT/CT imaging on estimation accuracy of left ventricular volumes and ejection fraction: A phantom study

摘要

The left ventricular (LV) volumes and ejection fraction (EF) are commonly calculated based on the partial volume effect (PVE) of gated single photon emission computerized tomography (G-SPECT) due to limited SPECT resolution. This study compares the accuracy of LV volume and EF quantification from a standard low resolution (LR) SPECT/CT system and a high resolution (HR) SPECT/CT system. Six custom-made cardiac phantoms each with a different inner and outer cylinder were used to simulate a wide range of LV volumes. The LV cavity (inner cylinder) was filled with water and the space between the inner and outer cylinders mimicking the LV myocardium was filled with Tc-99m solution. The cardiac phantom was inserted into a large cylindrical container filled with water. Images were acquired using a NaI/LR-SPECT/CT system (FWHM: 16 mm, GE Infinia) and a CZT/HR-SPECT/CT system (FWHM: 10 mm, GE D570c), and were reconstructed via the maximum likelihood expectation and maximization algorithm with CT-based attenuation correction. No photon scatter correction and post-filtering were applied to the images. The 6 sets of static SPECT images were aligned and assembled into an image sequence using our home-grown software to create 11-bin pseudo G-SPECT for a full cardiac cycle. Phantom volumes and EF were quantified using 4 different commercial software packages: Yale-Wackers-Liu CQ, Invia-Corridor4DM, Emory-Cardiac Toolbox and Cedars-Sinai-QGS. SPECT-quantified phantom volumes and EF were compared to the true phantom volumes and EF to evaluate the SPECT quantification accuracy. Quantification of LV volumes and EF from HR-SPECT using the 4 SPECT quantification software resulted in a significant underestimation of EF as compared to that from LR-SPECT, and also caused an erratic variation in the LV volume estimation. These findings were consistent for all 4 software. Current SPECT quantification software predominately based on the PVE may need to be adjusted to compensate for the improv- d spatial resolution associated with new HR-SPECT imaging.