Evaluation of 4D reconstruction methods for gated cardiac SPECT imaging in obese patients

The purpose of this study is to evaluate 4D reconstruction methods for the processing of gated cardiac single photon emission computed tomography (SPECT) images from obese patients. Gated SPECT on obese patients is extremely noisy and often clinically useless; it is hypothesized that 4D reconstruction methods may help. The methods compared are the ordered-subsets expectation-maximization (OS-EM) algorithm with a 3D Gaussian filter, OS-EM with a 3D Gaussian combined with a time-domain Butterworth filter, and the rescaled block-iterative maximum a posteriori (RBI-MAP) algorithm with Gibbs priors for spatial and time-domain smoothing. Clinical gated SPECT data were used to derive a table of Tc-99m tetrofosmin activity uptake ratios. Moderately and morbidly obese male and female phantom models were created for the 4D NURBS-based Cardiac Torso (NCAT) phantom, and mild and severe motion defects were generated in addition to a normal heart model. A blood pool phantom study enabled optimization of reconstruction parameters for the methods so they result in similar noise statistics in the heart. Poisson noise was added to the projection data (including the effects of detector response, attenuation and scatter) generated from the phantoms. The noisy phantom and patient projection data were reconstructed with the three methods, and imported onto the clinical workstations, to be analyzed with the Quantitative Gated SPECT (QGS) software. Quantitative parameters (chamber volumes) were recorded for the phantom and patient data. Statistical analysis led to the conclusion that OS-EM with 4D filtering was markedly different, a result confirmed in the normal phantom models, with better quantitation. Visually, RBI-MAP appeared to result in smoother, more realistic cardiac motion. A preference study was performed with four physicians who read the patient images using QGS and rated them on a 7-point scale to indicate which method most improved their confidence in the diagnoses. The one-way ANOVA showed no significant difference in preference for the processing methods. The conclusion is that the choice of reconstruction method may make more of a difference in patients with greater heart motion, and that the OS-EM method with 4D filtering may have an advantage over the other methods when it comes to LV chamber volume quantification