Scanner-Dependent Threshold Estimation of Wavelet Denoising for Small-Animal PET

Abstract

Reducing the noise associated with small-animal positron emission tomography (PET) images is an important and challenging task. Recently, several hybrid denoising techniques based on wavelet transform (WT) have been developed. However, these hybrid methods have complicated mathematical structures and require complex parameter estimations, and they therefore require a high level of manual intervention. Under such circumstances, good performance with respect to image quality using these new methods would only seem to be achievable with an increased computational burden. In this paper, we propose a novel wavelet denoising (WD) method. This method is based on scanner-dependent threshold estimation and the Visushrink method. The method provides a compromise between computational burden and image quality. The experimental results indicate that the proposed method is better than the Visushrink method. Compared with the Visushrink method, the proposed method provides good image quality at higher decomposition levels. In terms of usability and efficiency, the proposed method is better than the hybrid method. The proposed WD method also has several useful properties; therefore, it is possible that it might become an alternative solution to reducing the noise associated with small-animal PET images.