RT Journal Article
SR Electronic
T1 Effect of truncated overlapping projection data in a multi-pinhole brain SPECT system with temporal shuttering of apertures
JF Journal of Nuclear Medicine
JO J Nucl Med
FD Society of Nuclear Medicine
SP 39
OP 39
VO 60
IS supplement 1
A1 Navid Zeraatkar
A1 Benjamin Auer
A1 Kesava Kalluri
A1 Neil Momsen
A1 Lars Furenlid
A1 Phillip Kuo
A1 Michael King
YR 2019
UL http://jnm.snmjournals.org/content/60/supplement_1/39.abstract
AB 39Objectives: Brain SPECT imaging has clinical applications for diagnosis, prognosis, and treatment monitoring of neuro disease and brain tumors. We are designing and constructing a brain-dedicated SPECT system, AdaptiSPECT-C [1], with high resolution and sensitivity. It will be capable of both static and dynamic imaging with a stationary geometry employing multiple detectors. We have designed a multi-pinhole collimator with direct and oblique apertures for each detector to enhance the sampling capability of the system, hence reconstructed image quality. In this work, we study the effects of different levels of multiplexing (MUX) in the projection domain and acquisition (ASQ) time share schemes for the direct and oblique apertures using analytical simulations. Methods: The current design of the AdaptiSPECT-C consists of 23 hexagonal detectors in 3 rings: quasi-vertex (QV), middle, and caudal, arranged around a truncated spherical gantry. The collimator is composed of 5 pinhole apertures per detector in the caudal and middle rings, and 3 for the QV: one direct aperture at the center and 4 (2 for QV) off-centered oblique apertures. A pair of the oblique apertures is located along caudal-cranial direction (axial apertures) while the other pair is laterally positioned (lateral apertures). The QV-ring detectors do not have axial apertures. Each aperture can be independently opened/shuttered using a wireless module [2] we have developed. The shuttering allows data acquisitions of the direct and lateral/axial apertures in two separate temporal frames to avoid MUX data. However, different levels of MUX could potentially occur for each pair of the oblique apertures depending on their relative positions. Four MUX levels, from barely overlapping to the maximum physically allowed MUX, were simulated for each lateral/axial set of apertures. For all oblique apertures, projection images were truncated due to the finite size on the detectors. Four levels of total counts and 5 different combinations of ACQ time share, from only direct apertures to only lateral/axial apertures, were also modeled. An XCAT [3] head phantom with 123I-IMP perfusion distribution was used for all the simulations. Each simulation was performed with 5 noise realizations. Attenuation was also modeled in the simulations and the images were then reconstructed using a customized GPU-based MLEM algorithm. The normalized root mean square (NRMSE) was calculated from the reconstruction versus source distribution as the figure of merit for the reconstructed image quality. Results: The lowest NRMSE (averaged over noise realizations) over iterations was obtained for each data acquisition scenario. Based on NRMSE values, using oblique (axial or lateral) and direct apertures together led to better image quality than using only oblique apertures or only direct apertures. For almost all the studied cases, an ACQ time share of 50%-50% for direct-oblique apertures yielded to the lowest NRMSE. MUX levels of 16% and 27% (resulted in almost similar NRMSE values) for the lateral apertures and 26% for the axial apertures generated better NRMSE values while increasing MUX beyond these levels deteriorated the images. The additional apertures used with MUX increased the volumetric sensitivities considerably. Conclusions: We demonstrated that having additional oblique apertures with truncated and multiplexed projection images in the AdaptiSPECT-C geometry enhances image quality when incorporating the acquired data from both the direct (MUX-free and truncation-free) and oblique apertures in the reconstruction while no artefacts were seen in the region of interest. An optimization process was performed with a brain perfusion phantom and different levels of MUX, total counts, and ACQ time share of the apertures were studied. Research Support: National Institute of Biomedical Imaging and Bioengineering (NIBIB), National Institutes of Health, Grant No R01 EB022521.