Abstract
1386
Objectives In positron emission tomography (PET), images are degraded due to Poisson noise. The purpose of this study was to apply dual-tree complex wavelet transform and bivariate shrinkage method (CWTBS) on noise suppression of PET images.
Methods We used a mini deluxe phantom and a multi-line phantom to acquire images from the microPET® R4. After a transmission scan using a 68Ge rod source for attenuation correction, the transaxial images were reconstructed using the filtered back-projection algorithm to give a 256×256× 63 three-dimensional matrix with a pixel size of 0.423 mm ×0.423 mm × 1.121mm. Denoising techniques were applied on the reconstructed images. We also used the stationary wavelet transform with the soft-thresholding algorithm (SWST) to compare the results with our method using the coefficient of variation (CV), the contrast recovery coefficient (CRC), and the full width at half maximum (FWHM).
Results The following table shows the CVs, the CRCs, and the FWHMs of the various methods used.
Conclusions The selected parameters and image processing method in our study reduce the CV by 30%, maintain the CRC by 97.33%, while losing merely 3% of the FWHM. From the above results, it is evident that the dual-tree complex wavelet transform with bivariate shrinkage method takes the advantages on denoising and maintaining the contrast and spatial resolution of the images