Improved Quantitative SPECT Myocardial Uniformity of Myocardial Uptake Using a Half-Time, Iterative Resolution Recovery Algorithm
Authors: James A. Case, PhD; Jessica Jensen; Staci A. Courter; Paul Helmuth; Timothy M. Bateman
Introduction: Iterative resolution recovery methods have been used for reducing patient dose and improving laboratory efficiently. Validation techniques of these methods that depend on visual inspection of the images are prone to reader bias and hard to reproduce. We examined the quantitative uniformity of the myocardial counts following reconstruction using OSEM with resolution recovery (RR-OSEM) vs filtered backprojection (FBP) at full- and half-time acquisition times.
Methods: We examined twenty-six patients (21 male, BMI=29+/5) with a ECG-gated myocardial perfusion SPECT study acquired on a CardioMD (Philips) SPECT camera. Data was acquired in a 128×128 acquisition matrix (3.2 mm pixel size, 64 angles). Half-time data sets were created by summing only half of the gated frames. Raw datasets were reconstructed using FBP and RR-OSEM. Data sets were post-filtered using a 5th order, 0.4 Nyquist Butterworth filter. To reduce the influence of variations in perfusion uptake, the image was measured using a previously validated frequency domain masking technique. This technique performs a Fourier transform of the image and then removes the object by masking the lower frequency components. By performing an inverse Fourier transform on the masked frequency data, a “noise only” image can be made. The signal to noise ratio (SNR) was calculated as mean deviation of the myocardial ROI divided by the standard deviation (noise) in the same ROI in the frequency domain masked image.
Results: The signal to noise is summarized in Table 1. The square of the SNR in the half-time FBP reconstructions decreased by a factor of 0.56+/-0.11 when compared to the full-time FBP reconstruction. RR-OSEM half-time reconstructions were more stable, decreasing only by a factor of 0.87+/-0.18 vs. the full-time RR-OSEM reconstruction (p<10-15 FBP vs RR-OSEM).
Conclusions: The quantitative signal to noise was significantly higher with RR-OSEM vs conventional FBP reconstruction. Furthermore, changes in signal to noise were significantly less when going from full-time to half-time imaging.