Estimation of the PET Sensitivity and Spatial Resolution of the Human Dynamic NeuroChemical Connectome Scanner

Recent technological advances have led to high-sensitivity PET scanners, with increased fields of view, thousands of detectors, and billions of lines of response (LORs). The high sensitivity can be used to decrease radiotracer dose and/or acquisition time. Alternatively, it enables dynamic studies with unprecedented temporal precision. To preserve high spatial resolution, small section crystals and depth of interaction (DOI) corrections are mandatory, which increases by orders of magnitude the number of LORs. Cylindrical geometries chosen for high sensitivity whole-body scanners are suboptimal for dedicated brain PET scanners, for which spherical geometries would confer important advantages. Sensitivity can be further increased by proper consideration of triple (and beyond) coincidences, coming from photons which interact in more than one detector (inter-detector scatter). The Human Dynamic NeuroChemical Connectome scanner that will consist of a high spatio-temporal resolution (HSTR) BrainPET insert with a spherical geometry integrated with a 7T MRI scanner is currently being developed. Here, we present the results of Monte Carlo (MC) simulations aimed at assessing the HSTR-BrainPET scanner sensitivity before and after including triple-coincidences as well as phantom images reconstructed from simulated data using analytical and iterative approaches. The latter used a GPU-based iterative reconstruction algorithm capable of processing large datasets in either list mode or LOR-histograms. Resolution modeling was incorporated by means of a spatially variant point spread function, which was computed from realistic MC simulations considering DOI and other relevant effects. Our implementation enables high-resolution reconstructions within a few minutes, and demonstrates an achievable resolution better than 1.5 mm and sensitivity values higher than 25% in most of the relevant brain areas.