Keel-Edge Height Selection for Improved Multi-Pinhole ¹²³I Brain SPECT Imaging

Objectives: Given its excellent resolution versus sensitivity trade-off, multi-pinhole SPECT has become a powerful tool for clinical imaging of small human structures such as the brain [1]. Our research team is designing and constructing a next-generation multi-pinhole system, AdaptiSPECT-C, for quantitative brain imaging. In this context, keel-edge pinhole has proven to increase significantly attenuation of gamma rays through the edges of the pinhole aperture compared to the most commonly clinically used knife-edge profile [2-4]. In this work, we investigate the potential improvement in imaging performance of multiple keel-edge pinhole profiles as a function of keel height for AdaptiSPECT-C compared to a knife-edge collimation for ¹²³I-IMP brain perfusion. Methods: The prototype AdaptiSPECT-C system used herein is composed of 23 hexagonal detector modules hemi-spherically arranged along 3 rings. For modeling in GATE simulation (GS) [5], each of these modules is composed of 1.5 mm radius pinhole and a 1 cm thick NaI(Tl) crystal with a 5 cm thick back-scattering compartment, which was considered to simulate ¹²³I down-scatter interactions. Multiple keel-edge heights, corresponding to 0.0 (knife edge), 0.375, 0.75, 1.0, 1.125, 1.5, 1.875, and 2.25 mm were studied. We evaluated the volumetric sensitivity and relative amount of collimator penetration for a 15% energy window centered at 159 keV in simulated projections of a 21 cm diameter spherical source (e.g. corresponding to the system’s volume of interest) centered at the focal point of the pinholes. For reconstruction, an approach developed in our group was employed for modeling using GS the system response and especially collimator penetration into the system matrix [6,7] for the knife and the keel-edge designs. An XCAT [8] brain phantom with source distribution for the perfusion imaging agent ¹²³I-IMP was simulated using the pinhole designs. Projection were acquired considering two scenarios: noise free (S1), and equal imaging time comparison for a realistic clinical scan time (e.g. 30 min [9,10]) (S2). Reconstructions were performed with a customized 3D-MLEM software into images of 120³ voxels of (2 mm)³. The reconstructed images were then compared to the ground truth image in terms of the normalized root mean squared error (NRMSE) and activity recovery (%AR) for selected three-dimensional brain regions.

Results: A keel-edge height of 0.375-0.75 mm represents the best choice leading to a significant reduction of the amount of penetration (up to ~50%) at the expense of sensitivity (-20%) compared to a knife-edge profile. Visually, for all scenarios, the use of such a keel-edge profile leads to better separation of the brain structures, especially the caudate and the putamen. When sensitivity is not taken into account (e.g. noise free scenario), increasing the keel height improves NRMSE results. For an equal imaging time comparison, lowest NRMSE values are achieved for a 0.375-0.75 mm keel-height. A further keel-height increase degrades the NRMSE results due to significant loss of counts compared to knife-edge design. A 0.75 mm keel height leads on average to the best %ARs (e.g. closest value to 100%), especially for the striatum and putamen. For cortex and cerebellum regions, %ARs are comparable with those obtained for a knife-edge design. Conclusion: In this work, we demonstrated that the use of a ~0.75 mm height keel-edge profile for AdaptiSPECT-C incorporating 1.5 mm radius pinholes leads to superior imaging performance compared to knife-edge collimation for clinical ¹²³I brain perfusion imaging. A range of aperture radii from 0.5 to 3.5 mm for each design have been investigated and will be shown at the time of the conference. We are currently working on performing a numerical-observer task-performance study of defect-detection in perfusion. Research Support: National Institute of Biomedical Imaging and Bioengineering (NIBIB), National Institutes of Health, Grant No R01 EB022521.