Noise reduction of parametric images of myocardial blood flow by filtering H215O dynamic PET images using wavelet transform

Dynamic PET images are useful for the measurement of physiological functions. However, signal to noise ratio of H₂¹⁵O dynamic PET images is poor, thus filtering is necessary. It is well known that discrete wavelet transform saves detailed information in high frequency while diminishing noise. Recently, two methods to generate parametric images of myocardial blood flow using H₂¹⁵O dynamic PET images have been suggested by our group, but the signal to noise ratio of the suggested parametric images has a room for improvement by applying appropriate temporal or spatial filters to raw dynamic images. Thus, in this study, we applied wavelet transform to reduce the statistical noise in time-activity curve in each pixel prior to generating parametric images of myocardial blood flow and related parameters, and compared the image quality of parametric images with and without wavelet filtering. Wavelet denoising applied to raw dynamic images prior to generating parametric images, removed noise in time-activity curves and, as a result, increased image quality of parametric images without the degradation of spatial resolution.