Study of multispectral frame-by-frame convolution scatter correction in high resolution PET

High resolution PET scanners based on small individual detectors have a low sensitivity resulting from increased inter-detector scatter. It has been postulated that this limitation can be overcome by using enlarged discrimination windows to include more low-energy events and by developing more efficient energy-dependent methods to correct for scatter. In this work, we investigate one such method based on the frame-by-frame scatter correction of multispectral data. Images acquired in the conventional (344-658 keV), broad (129-658 keV) and multispectral (129-658 keV) window modes were processed by stationary and nonstationary convolution-restoration scatter correction methods. Broad and multispectral window acquisition improved counting efficiency relative to the conventional window. The degradation of image quality due to added scatter can be mostly recovered by the subtraction-restoration scatter correction. However, the multispectral frame-by-frame method was found to be more sensitive to the effects of the nonstationary scatter response functions and its performance was not as good as that of the broad window. It is concluded that more sophisticated scatter degradation models and correction methods need to be established to take full advantage of multispectral data.