Skip to Main Content
Gated blood pool imaging (GBP) is well known as an accurate method for evaluation of the main parameters of cardiac function. However, the interpretation of the raw data is difficult because of insufficient images quality due to the approximations applied during the acquisition and to the fact that GBP is done in vivo (heart motion or gamma rays diffusion). Thus images denoising and/or smoothing must be performed. At the same time, image processing must be performed with a maximum of care and precision. Previous studies have been showed excellent performances of wavelet transform for denoising and compression of signals and images. In our department, we routinely use Karhunen-Loeve transform (KLT) to process GBP images. In this study, we revised our previous results concerning the accuracy of spatio-temporal smoothing of image sequences by using of KLT and compared the features of wavelets and of KLT for GBP images denoising. GBP studies using albumin marked with 99mTc (12 MBq/kg) were performed in 30 patients at rest. The image processing was performed in MATLAB version 5.3 on standard PC (PIII-733, 256 Mb RAM). We performed visual analysis of resulting images, analysis of mean image intensities, analysis of gradient images, spatial Fourier analysis and analysis of time-activity curves. The comparison of smoothed images showed that both methods offer adequate quality of images denoising and facilitate the detection of contours. However, the analysis of time-activity curves showed the superiority of KLT; we observed only slight modifications of time-activity curves after the smoothing. In conclusion, although our results showed good accuracy of both methods-wavelets and KLT-for spatial smoothing, only KLT offers possibility of adequate spatial and temporal smoothing of GBP images.