Chapter 33 - The Effect of Spatial Correlation on the Quantification in Positron Emission Tomography

Abstract

In positron emission tomography, the values in adjacent parts of reconstructed images of radioactivity distributions are correlated. These correlations are examined by phantom measurements. The corresponding correlations in the parametric maps, derived from a time series of radioactivity maps, were investigated analytically and by simulation. The correlation is found to depend strongly on the distance between the pixels and on the cutoff frequency of the filter function used in the image reconstruction. With 3D data acquisition, there is also a considerable axial correlation, which is found to be largest between the outer planes and to increase with distance from the camera axis. The effect of correlation on the standard error of the mean (SEM) of the pixel values in a region is found to be considerable and depend on both the size and shape of the region. For kinetic models, linear in the parameters, it is shown that the correlation coefficient between parameter values in two pixels is equal to or smaller than the correlation coefficient between the corresponding radioactivity values. Equality holds only when the ratios between the errors in two pixels are the same in all time intervals. The simulations show that parameters entering nonlinearly in the model equation tend to have smaller correlation coefficient than parameters entering linearly, which have almost the same correlation coefficient as the underlying radioactivity data.