Implementation of noise reduction for PET using hybrid nonlinear wavelet shrinkage method

Positron emission tomography (PET) imaging provides the functional information and precise physiological uptake of radioactivity in a patient’s body. But the photon level of PET system has fewer level, the shortcoming of PET is low SNR. The quantum noise in emission data can have a significant influence on the statistical uncertainty of PET measurements. The reduction of the quantum noise in images is an important issue. Conventional post-processing is not completely suitable, since it leads to loss of the spatial resolution while reducing noise. In this study, we present a method for denoising the quantum noise in images. Our proposed methods is to use a hybrid wavelet shrinkage processing in 3D sinogram(bin,angle,slice) domain. One is applied to axial(bin,angle) space to reduce detector dependence noise element, the other is applied to 3D space to reduce quantum noise. We optimized three parameters in each wavelet processing. The efficacy of the filtering technique was shown on a derenzo phantom. Finally, we also performed human FDG studies to evaluate clinical performance. As result, we have found that the optimized parameters and reconstructed image without post filtering in proposed study reduce COV while maintaining average and spatial resolution. We therefore conclude that wavelet de-noising technique is an effective tool to process PET images and thus to provide better image quality. We expect that our study can provide benefits to clinical applications of PET images.