Evaluation of Rotating Multi-Segment Slant-Hole as Compared to Parallel-Hole Collimation for Myocardial Perfusion SPECT



Liu, Chi; Xu, Jingyan; Tsui, Benjamin M.W.



The purpose of this study is to evaluate the imaging characteristics of rotating 4-segment slant-hole (R4SSH) as compared to parallel-hole (PH) collimation for myocardial perfusion SPECT (MPS) imaging in terms of defect contrast, noise level, contrast-to-noise ratio (CNR), and defect detection using the channelized Hotelling observer (CHO).


A R4SSH collimator with a resolution of 9mm at 15cm, the distance to the center of common-volume-of-view (CVOV), and a GE low-energy high-resolution PH collimator with a resolution of 11mm at 22cm were used in this study. For R4SSH SPECT, the heart was placed at the center of the CVOV and projection data were acquired using 3 camera positions, each with 12 rotation stops. For PH SPECT, 64 projections over 180 degree with a radius-of-rotation of 22cm were acquired. The simulated noise-free projections were obtained from the 3D NCAT phantom using analytical projectors for both collimations that include the effects of attenuation and collimator-detector response. They were then scaled and Poisson noise was added to simulate noisy projections obtained from R4SSH and PH SPECT with the same acquisition time. Iterative OS-EM algorithms with 4 subsets with and without attenuation compensation (AC) for both collimations were applied. Defect contrast was defined as the ratio of intensities at the center of the myocardial defect and the normal myocardium in the noise-free reconstructed images. Noise level was measured by the normalized standard deviation (NSD) at normal myocardium obtained from 30 noise realizations. The CNR was calculated as the ratio of the defect contrast and the NSD. The CHO methodology and receiver operating characteristic (ROC) analysis were applied to the short-axis image slices that contained the defect centroids to determine the optimized number of iterations for the OS-EM reconstruction method and the cutoff frequency of 3D Butterworth post-filter in terms of area under the ROC curve (AUC) for the myocardium defect detection task.


When compared to PH SPECT, R4SSH reconstructed images were found to have substantially lower noise level and higher CNR. For both R4SSH and PH SPECT, AUCs are more sensitive to the cutoff frequency than to the number of iterations. For both with and without AC, using the optimal combinations of number of iterations and cutoff frequency, the maximum AUCs of R4SSH SPECT are significantly higher than those of PH SPECT with p-values <0.01.



Due to higher detection efficiency and better spatial resulution, R4SSH SPECT provides MPS images with lower noise and higher CNR than those of PH SPECT. Using the same acquisition time and appropriate imaging parameters, R4SSH SPECT was found to have significantly higher myocardial defect detectability than PH SPECT.



This work is supported by the National Institutes of Heath under Grant R01-EB001983