Abstract – The ECAT HRRT (High Resolution Research Tomograph, CPS Innovations, Knoxville, TN, U.S.A.) is a 3D-only brain positron tomograph made of height flat panels using LSO and LYSO scintillators. This dual-layer octagonal geometry allows the scanner to measure depth-of-interaction, preserving the good spatial resolution toward the edge of the transverse field- of-view. With the objective of optimizing protocols and acquisition parameters, we used the GATE (Geant4 Application for Tomographic Emission) Monte Carlo simulation platform to build a realistic model of this scanner. The aim of this paper is to compare the performance of the model against the real system. In order to remain the comparison being independent on the reconstruction scheme used, we generated the simulation data as the HRRT native 64-bit list-mode format before software histogramming and rebinning into sinograms. We performed for the comparison three sets of experiments : intrinsic spatial resolution, scatter fraction and count rates. Real and simulated data were found to be in good agreement. To illustrate the use of this model we evaluated the crystal thickness influence for a single layer HRRT on spatial resolution and sensitivity. I. INTRODUCTION Positron emission tomography represents today an essential role in modern medicine for both diagnostic and treatment purposes. Growing requirements in velocity and precision imply the optimization of acquisition parameters and protocols. Monte Carlo simulations are essential tool to assist these developments, improving data analysis and image quantification. GATE [1] (Geant4 Application for Tomographic Emission) is such a simulation platform, based on the Geant4 toolkit package. Dedicated to nuclear medical imaging, GATE allows to describe time-dependent phenomena such as detector movement or source decay kinetics, thus allowing to simulate time curves under realistic acquisition conditions. We used this tool to build a realistic model of the ECAT HRRT [2] (High Resolution Research Tomograph, CPS Innovations, Knoxville, TN, U.S.A) 3D-only dedicated brain scanner. In this paper, we compare the performance of the simulator against the HRRT scanner in order to validate the use of the model for system optimization. Manuscript received October 29, 2004. F. Bataille, C. Comtat, S. Jan, and R. Trébossen are with the Frédéric Joliot Hospital Facility (SHFJ), CEA, Orsay, France. II. METHODS A. HRRT geometry The HRRT is a brain positron tomograph made of 8 detector heads arranged in an octagon. Each head consists of 9 × 13 blocks of 8 × 8 dual-layer 2.1 × 2.1 × (10 + 10) mm 3 LSO- LYSO crystals, in coincidence with five opposing heads. The head-to-head distance is 469 mm and the axial field-of-view is 253.5 mm. A simulation of this geometry is represented in Fig. 1. The phoswitch configuration allows the scanner to measure depth of interaction, preserving the good spatial resolution toward the edge of the transverse field-of-view. The data are acquired in a 64-bit list-mode format and then software histogramed into sinograms. Fig. 1: HRRT geometry simulated with GATE, where the LSO layer is green, the LYSO is yellow and the tungsten end-shielding is blue. B. GATE configuration For each detected photon, the hit position corresponds to the crystal with the maximum energy deposition, assuming an energy resolution varying between 20 % and 30 % for each crystal. GATE capability to synchronize all time-dependent components allows for the explicit simulation of coincidence events. As for the real acquisitions, we used a 6 ns coincidence time window. We used the ROOT [3] output provided by GATE to exploit the simulation results. This structured list- mode file contains event-by-event characteristics for three acquisition levels : crystal hits, singles and coincidences. We developed a macro to convert this output into the native 64-bit list-mode required by the HRRT embedded software. This method allowed us to reconstruct real and simulated data with the same schemes. F. Bataille, C. Comtat, Member, IEEE, S. Jan, and R. Trébossen Monte Carlo Simulation for the ECAT HRRT using GATE 0-7803-8701-5/04/$20.00 (C) 2004 IEEE