A knowledge-based image smoothing technique for dynamic PET studies

Abstract

Many techniques have been proposed to reduce image noise in dynamic positron emission tomography (PET) imaging. However, these smoothing methods are usually based on the spatial domain and local statistical properties. Smoothing algorithms specifically designed for dynamic image data have not previously been investigated in detail. We present a knowledge-based smoothing technique that aims to diminish the noise and improve the quality of the dynamic images. By taking advantage of domain specific physiological kinetic knowledge, this technique can provide dynamic images with high noise reduction while preserving edges and subtle details.