Acquisition strategies of a dual head rotating 4-Segment Slant-Hole (R4SSH) SPECT System for Improved Myocardial Perfusion SPECT Imaging

The goal was to evaluate different acquisition strategies (AS) of a dual-head (DH) Rotating 4-Segment Slant-Hole (R4SSH) SPECT system for improved myocardial perfusion (MP) SPECT imaging. Two identical R4SSH collimators were designed to fit on the two large-field-of-view (LFOV) cameras of a clinical DH SPECT system. Both R4SSH collimators had the same system spatial resolution of 1.15cm at the center of the common-volume-of-view (CVOV) which was 21.5cm from the collimator faces. They were positioned off center of the LFOV camera such that their CVOVs coincided with each other resulting in a 82.5° angle between the two LFOV cameras. Multiple projections of the heart were acquired by positioning the heart at the center of the CVOVs of both R4SSH collimators and rotating them. Additional projections can be obtained by rotating the DH SPECT gantry to another position while keeping the heart at the centers of the CVOVs. To evaluate the AS of the DH SPECT system with two R4SSH collimators, realistic projection datasets from the 3D XCAT phantom modeling Tc-99m MIBI uptake distribution were simulated using an analytical projector modeling attenuation and collimator response function. Six different AS were generated by placing the dual cameras at different orientations with respect to the heart. At each SPECT gantry position, the R4SSH collimators rotated in 12 steps over 90°. The simulation was repeated using a normal and an enlarged heart. A 3D R4SSH OS-EM method was used for image reconstruction. The six different AS were evaluated by comparing the reconstructed images with the corresponding XCAT phantom slices. Our results showed that the difference between the R4SSH SPECT images and the XCAT phantom was greatest when the long axis of the left ventricle (LV) was pointed to the middle of the dual cameras and was least when it was perpendicular to the face of either camera. In the latter case, the quality of the R4SSH images was similar to that obtained fr- m the complete projection dataset with 4 camera positions. For practical OS-EM update numbers, the normalized mean squared error (NMSE) for 3 of the 5 AS using 2 camera positions was lower than that using 4 camera positions. These results show that by optimizing the DH SPECT gantry position, a stationary DH SPECT system fitted with 2 R4SSH collimators can obtain similar quality MP SPECT images as that obtained by an additional system rotational position.