TY - JOUR T1 - Correction of random events for a dual-head PEM detector JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 1982 LP - 1982 VL - 52 IS - supplement 1 AU - Cláudia Ferreira AU - Ricardo Bugalho AU - Catarina Ortigão AU - Mónica Martins AU - Nuno Matela AU - João Varela AU - Pedro Almeida Y1 - 2011/05/01 UR - http://jnm.snmjournals.org/content/52/supplement_1/1982.abstract N2 - 1982 Objectives In PEM, the correction of random events is important due to the contribution from high uptake organs (e.g heart). Their effects can be reduced by narrowing the coincidence time window together with the use of an efficient randoms correction method like the delayed coincidence subtraction method (DW). In this paper, we focus on the correction of random events for the Clear-PEM detector, a dual-head breast PET system. Methods We tested DW with the variance reduction on randoms (VRR) method proposed for the HRRT. Tests were done for a dual-head PEM scanner using simulated (no attenuation or scatter) and real data with significantly different statistics. Simulated uniform cylinders with different activity concentrations were used to create simulated true and random events. Real data was obtained with two Ge-68 sources one of which was located outside the camera's FOV. A real breast exam was also used to evaluate the method's performance. The signal-to-noise ratios (SNR) on images were obtained for a total of thirty ROIs drawn over interest regions and background. Data were reconstructed using a modified version of OSEM3D suitable for plane detectors. The VRR method results were compared to the use of DW with no VRR. Results In simulated data, the use of VRR increases the data SNR values by a factor of 1.5 when compared to the use of no VRR. Improvements obtained with the use of VRR in real data ranged from 4% (clinical data) to 13% (Ge-68 phantom data). Conclusions The use of VRR techniques on a DW method seems adequate to correct randoms in dual-head PEM data. Its use results in significant increases of SNR in reconstructed images. The performance of the VRR method may depend on object statistics, which were different on simulated and real data. Research Support The Clear-PEM project is financed by AdI and FCT/POSI - Portugal. The work of C. S. Ferreira, N. Matela and R. Bugalho was supported by FCT (Portugal)- Grants SFRH/BD/47448/2008, SFRH/BPD/29696/2006 and SFRH/BD/66008/2009 ER -