Detailed simulations of a full-body RPC-PET scanner M. Couceiro1,2, P. Crespo1, R. Ferreira Marques1,3, P. Fonte1,2 1 LIP, Laboratório de Instrumentação e Física Experimental de Partículas, Coimbra, Portugal 2 ISEC, Instituto Superior de Engenharia de Coimbra, Coimbra, Portugal 3 Departamento de Física da Universidade de Coimbra, Coimbra, Portugal

Introduction to PET Brief PET overview PET is a medical imaging technique, aimed to study functional processes An appropriate molecule, labelled with a positron emitter radioisotope, is injected in the patient When a decay occurs, a positron is emitted, which loses energy in several collisions, combining then with an electron from the medium Since the positron is the electron antimatter, an annihilation occurs, resulting in two 511 keV photons emitted in opposite directions Annihilation photons, detected in an appropriate time window (4 to 12 ns), are recorded, defining a Line Of Response (LOR) Image is reconstructed from the acquired LOR In PET we are interested in detecting photons from positron-electron annihilation. The image is reconstructed from acquired LOR.

Introduction to PET True Scattered Random Brief PET overview Three types of coincidence events can occur True coincidences: photons from a single decay leave the patient without suffering interactions Scattered coincidences: one or both photons from a single decay interact in the patient, losing energy, and changing their initial direction Random coincidences: photons from different decays detected in coincidence True Scattered Random False lines of response  must be rejected Three types of coincidence events can occur: True, scattered and random coincidences Scattered and random coincidences define false lines of response, and must be rejected

Introduction to PET Brief PET overview State of the art PET scanners Scintillation crystal detectors with high efficiency for 511 keV photons and good energy resolution for scatter rejection Reduced Axial Field Of View Several bed positions to obtain a full body image Increased injected activity Discontinuous uptake signal Full body PET scanner Full Axial Field Of View Single bed position to obtain a full body image Reduced injected activity Continuous uptake signal Too expensive with crystal detectors RPC may be a suitable detector for full body PET scanners [D.B.Crosetto, 2000] State of the art PET scanner use scintillation crystal detectors with high efficiency for 511 keV photons and good energy resolution for scatter rejection. However they have a reduced axial field of view, needing several bed positions to obtain a full body image, increased injected activity and introducing discontinuities in the uptake signal. A full body PET scanner with an extended Axial Field of View, could address all these problems. However, it would be too expensive with crystal detectors. RPC may be a suitable detector for this application.

Resistive Plate Chamber Basics XY readout plane Y-strips RC passive network X-strips Resistive Cathode Resistive Anode At least one resistive electrode E Time signal HV Sensitive Area (precise small Gas Gap ~300 µm) Photon e- [A. Blanco et al., 2003] For 511 keV photons, and commonly used materials Good timing resolution of 300 ps FWHM for photon pairs No energy resolution < 0.4% efficiency per gap for singles For 511 keV photons, and commonly used materials, single gap RPC has a good timing resolution of 300 ps FWHM for photon pairs, no energy resolution, and very low detection efficiencies, which can be compensated by stacking several plates Stacked RPCs .......... e-

Transaxial Section of a Scanner Detection Wall RPC-PET Simulations – Scanner geometry Scanner Geometry 1151.6 mm 2408 mm Transaxial Section of a Scanner Detection Wall 143.6 mm 1008 mm For the simulations the scanner was defined has four detection walls, each containing a stack of 20 RPC detectors, with 10 gaps each Top Glass Stack Bottom Glass Stack RPC Detector 0.20 mm Borosilicate Glass Top Y Electrode 0.35 mm Nylon Spacers X Electrode Bottom Y Electrode

RPC-PET Data processing – Dead time processing Each detector, has 10 independent transaxial readout sections (total of 800 for the scanner) Non-paralyzable dead time for time signals Paralyzable dead time for position signals Coarse axial position readout Fine transaxial position readout Time readout Fine axial position readout Axial Direction Transaxial Direction [Patent pending] Each detector was considered to have 10 independent readout sections in the transaxial direction

RPC-PET Data processing – Dead time processing (readout section) 1 Hits 2 3 Non-paralyzable dead time model applied to all simulation hits Readout generated time events Time Event (te1, xe1, ye1, ze1) (te2, xe2, ye2, ze2) Final Event Paralyzable dead time model applied to all valid time events 1 1 Time Event 2 3 2 3 2+3 Fine position (3.44 mm binning in the radial direction, and 2 mm binning in the transaxial and axial directions) Both events can be rejected or accepted with a coarse position (3.44 mm binning in the radial direction, 3 cm binning in the transaxial direction and 1 cm  Gaussian blur in the axial direction) A non-paralyzable dead model is first applied to all simulation hits to generate valid time events, which are then processed with a paralyzable dead time model. In case of pileup, both events can be rejected, or accepted with a coarse position.

RPC-PET Data processing – Coincidence sorter  = 5 ns 1 Event 2 3 Single time window coincidence sorter 4 5 Single coincidence Multiple coincidence  = 5 ns  1 Event 2 3 Multiple time window coincidence sorter 4 5 Single coincidences Multiple coincidences (1,2) (4,5) (2,3,4) (3,4,5) For the coincidences, single and multiple time window sorters were used

RPC-PET Results – Scatter fraction (NEMA 2001) Plotting the shifted sinogram projection against the radial position shows that detector scatter is responsible for long-range diffusion. Detector scatter is responsible for some long-range diffusion

RPC-PET Results – Time-Space patterns for coincidence ? Best trigger ? Plotting the time difference between coincidence photons against distance between them, for single, multiple and all possible coincidences, it was found that coincidences lying above the light speed line are due to scatter in the scanner. Besides, true coincidences are confined to this region, and multiple coincidences also contain true events. What is the best trigger?

RPC-PET Results – Tested trigger strategies Time-Space rejection plus Geometric TOF rejection + 300 ps Geometric TOF rejection Bore rejection Geometric rejection We tested several options: bore rejection, geometric rejection, time-space rejection and geometric TOF rejection with and without a 300 ps tolerance.

RPC-PET Results – Scatter fraction (NEMA 2001) Single time window coincidence sorter Multiple time window coincidence sorter The scatter fraction obtained is lower for the geometric tof trigger, and single time window coincidence sorter. However, with the multiple time window coincidence sorter there is only a small excess of scatter.

0.0 s dead time for position signal RPC-PET Results – Count rates and NECR, with position pileup rejection (NEMA 2001) Geometric rejection Geometric TOF rejection 0.0 s dead time for position signal Plotting the count rates and NECR with position pileup rejection, for zero micro seconds position readout dead time, it is clear that geometric tof rejection drastically reduces the random count rate, reducing also the scatter count rate, and essentially maintaining the true count rate. As a consequence, the NECR increases. This behavior is independent of the dead time for position signals.

0.5 s dead time for position signal RPC-PET Results – Count rates and NECR, with position pileup rejection (NEMA 2001) Geometric rejection Geometric TOF rejection 0.5 s dead time for position signal

1.0 s dead time for position signal RPC-PET Results – Count rates and NECR, with position pileup rejection (NEMA 2001) Geometric rejection Geometric TOF rejection 1.0 s dead time for position signal

3.0 s dead time for position signal RPC-PET Results – Count rates and NECR, with position pileup rejection (NEMA 2001) 3.0 s dead time for position signal Geometric rejection Geometric TOF rejection

Independent of dead time for position signal RPC-PET Results – Count rates and NECR, with coarse position (NEMA 2001) Independent of dead time for position signal Geometric rejection When pileup occurs for the position signals, and the events are accepted with a coarse position, the results are independent of dead time, and equal to those previously shown for a dead time of zero micro seconds Geometric TOF rejection

RPC-PET Results – NECR for single coincidence pairs (NEMA 2001) This plot shows the NECR obtained for single coincidence pairs, with the tested configurations for the coincidence sorters and position pileup policies considered, along with a plot of the GEMINI TF scanner.

RPC-PET Results – NECR for multiple coincidence pairs (NEMA 2001) The same only for the multiple coincidences, showing that they can positively contribute to the total NECR

RPC-PET Results – NECR for all possible coincidence pairs (NEMA 2001) Combining the single and multiple coincidences, NECR of RPC-PET is expected to be higher than that of one of the best tomographs currently available

RPC-PET Results – Spatial resolution (NEMA 2001) DOI + 2.0 mm Binning Mean = 2.1 mm DOI + 1.0 mm Binning Mean = 1.4 mm Only DOI Mean = 0.8 mm [M.Couceiro et al, 2012] The spatial resolution computed at low count rates was found to depend on detector binning, being better then that presented by comercial PET scanners

Conclusions Disadvantages in comparison to crystal based detectors Much smaller detection efficiency (20% to 50%) No energy resolution (although energy sensitivity) Previous simulations suggested that RPC-PET will have a significant improvement in sensitivity over current technology Current simulations suggest that RPC-PET has higher scatter fractions than those present in current crystal based PET scanners, which however do not compromise system performance, as measured by the NEMA NU-2 2001 However, it is still necessary to test the spatial resolution at high count rates, and image quality as stated in NEMA NU-2 2001 protocol to conclude if coarse position policy can effectively be used to further increase NECR

Acknowledgments Laboratory for Advanced Computation (LCA) of the University of Coimbra, Portugal, for a very generous gift of computing time (300,000 hours) Miguel Oliveira, of the LIP computing center, for his prompt response and availability to solve computational problems, including storage ones