1. 0 A Fast Time Incorporating Monte-Carlo Simulation of Wire Chamber Based Small Animal PET Scanners for Detector Scatter Correction M. Dawood 1, Don Vernekohl 2, K. P. Schäfers 1 1 European Institute for Molecular Imaging, University of Münster 2 Institute for Mathematics, University of Münster

2. 1 Scatter correction scheme Use measured (trues+scatter) data to simulate scatter Subtract simulated scatter from data Reconstrut correced data

3. 2 HIghgDensity AvalanchingvCathode Detector quadHIDAC 16 cm 28 cm

4. 3 quadHIDAC converter Photon Electrons Lead Insulation 0.4 mm Holes Converter for detection of the gamma photons

5. 4 quadHIDAC module Anode wires Y-Cathode tracks X-Cathode tracks Converter A Converter B Photon

6. 5 Detector scatter Detector scatter : compton & elastic scatter Detector Scatter:~ 40 % of all events Attenuated events:~ 8 % (mouse)

7. 6 Remodeling the converters Lead density ρNew lead density ρ‘ 0.0113 g/mm 3 0.0018 g/mm 3 Lead density adjustment for change in volume and surface

8. 7 Fast Monte-Carlo simulation Physics – All events are generated at once – Propogation until detectors are reached

9. 8 Fast Monte-Carlo simulation Scattered True Undetected Physics – Simulation of photo, compton, and thompson events

10. 9 Time component Time stamps assigned to all events Time windows – Coincidence window 40 ns – Dead time 160 ns – Coincidence deadtime 400 ns

11. 10 Time component Even and odd converters are grouped Dead time is dependent on the even and odd converters

12. 11 Dead time effects Trues vs Activity curve shows the effect of deadtime

13. 12 Scatter distribution SimulationGeant Transversal cross section of a simulated point source

14. 13 Scatter distribution on the detectors X-Coordinate # of events

15. 14 Point source TruesTrues + ScatterCorrected EMrecon algorithm and scatter correction see: M13-7, Scatter and Random Corrections…, Vernekohl et al. Today, 16:30-18:30 Hall B2 10 Million annihilations simulated at the center of the FOV

16. 15 Scatter estimate in a mouse Scatter estimated from a mouse dataset GEANT Simulation Coronal sliceSagital slice

17. 16 Mouse data Uncorrected Mouse Data GEANT CorrectedSimulation Corrected

18. 17 Computational efficiency The simulation in „Matlab“ – 1 Million positron annihilations ~ 1.2 hour vs. ~ 8 hours with GEANT (single processor) – Parallel computation implemented ~ 10 Min with 12 processors

19. 18 Thank you for your attention!