De-noising of SPECT images via optimal thresholding by wavelets

Single photon emission computed tomography (SPECT) imaging provides functional information and precise physiological uptake of radioactivity in a patient´s body. Although SPECT imaging is considered to be highly useful in oncology, but the low signal to noise ratio (SNR) caused by photon noise, introduces considerable compromise in image quality and reduction of diagnostic accuracy. It is necessary to apply appropriate noise reduction algorithm to improve the quality of acquired images. In this paper we have used wavelet based denoising in which PSNAR criteria were utilized to arrive at an optimum thresholding of the coefficient at wavelet domain. We have used SIMIND software for simulation of SPECT images and generation of images using cylindrical jaszak phantom. The images were acquired using one million counts of 64times64 matrix size. In this research, simulated images were utilized to construct data dependent optimum threshold level of wavelet coefficients We have compared the results of our thresholding scheme with those obtained by some of commonly used standard denoising schemes in which we show the use of commonly used wavelet-based denoising leads to an inferior noise reduction results as compared with our optimally searched and derived thresholding.