Frequency selective signal extrapolation for compensation of missing data in sinograms

Proceeding of: 2008 IEEE Nuclear Science Symposium Conference Record (NSS '08), Dresden, Germany, 19-25 Oct. 2008We present a method to compensate for missing projection data in positron emission tomography (PET), which may result from gaps between adjacent detectors or from malfunctioning detectors. To prevent artifacts in the reconstruction when using Fourier rebinning (FORE) or analytical reconstruction algorithms, filling the data gaps is required. This new approach for sinogram data interpolation prior to reconstruction is based on a simple iterative freq uency selective signal extrapolation method initially proposed and successfully applied for error concealment in image and video communication. In this work the method has been improved taking advantage of well known restrictions in the allowed region of frequencies of the sinograms. We compare quantitatively the results of this technique with previously proposed gap filling procedures in both the sinograms and in their reconstructed images. The proposed technique outperforms those methods without requiring too much computational time.This work was supported in part by the MEC (FPA2007-07393), CDTEAM (CENIT-Ingenio 2010) Ministerio de Industria, Spain, CPAN (Consolider-Ingenio 2010) CSPD-2007-00042 projects, and the RECAVA-RETIC network. Vicente acknowledges support from a CSIC-JAE predoctoral fellowship