Abstract
2396
Objectives Using collimation in PET can achieve spatial resolution beyond the limit of detector resolution. We use simulations and experiments to characterize/optimize the collimator and guide the design of a proposed PET collimator.
Methods We design trapezoidal tungsten collimator septa to mask half of each crystal in a PET scanner to detect lines of response (LORs) within the unmasked portion of the crystals. Since each crystal can be conceptually split into 2 sub-crystals, each pair can measure 4 finer LORs with 4 collimator configurations. By dividing the collimator into 3 segments, we can achieve complete LOR sampling inside a triangular region containing the field of view (FOV) with 4 sampling configurations. We developed a PET-collimator simulator based on 3D ray tracing incorporating the collimator’s geometric parameters. We also performed point-source experiments with trapezoidal tungsten pieces inside A-PET, a small-animal PET scanner. Crystal responses and LOR resolutions with different configurations were measured and compared with simulations. Based on A-PET, collimators with feasible parameters were simulated inside a detector ring of 20.7 cm diameter with 278 ideal crystals of 2.34 mm width. Two metrics, LOR resolution and average sensitivity over a 15 mm FOV, were used to characterize collimator performance. LOR resolution was determined by measuring the FWHM of the coincidence response function of a point source near the FOV’s center. Sensitivity, including geometric and penetration, was averaged over the FOV and 4 sampling configurations. We optimized the collimator by maximizing this average sensitivity with fixed LOR resolution within feasible geometric parameters.
Results After optimizing, LOR resolution of 0.6 mm with collimation can be achieved with relative sensitivity of 22.7% compared to uncollimated resolution of 1.17 mm.
Conclusions Collimation can dramatically increase sampling and improve spatial resolution but with reduced sensitivity. Collimated PET can improve overall image quality and quantification, which may have significant implications in clinical imaging where high resolution is demanded.
Research Support This work is supported by NIH grant R21-EB-12158