UNLABELLED: Although removal of object scatter has been shown to improve both contrast and quantitation accuracy, subtraction of detector scatter leads to marginal contrast enhancement and negligible resolution recovery at the expense of reduced sensitivity and increased statistical noise. Since detector scatter has correct information about radioactivity but slightly erroneous information about source location, we suggest that this component should be restored to preserve sensitivity and improve resolution. METHODS: A scatter correction model that consecutively removes object scatter and restores detector scatter is proposed. The scatter components are processed in the spatial domain using nonstationary scatter kernels. The detector scatter restoration kernel is obtained by piecewise inversion in the Fourier space. The model was tested using line source and hot spot phantom measurements. RESULTS: Object scatter subtraction increased contrast substantively with no effect on resolution. Detector scatter restoration recovered resolution almost completely with modest contrast enhancement in small lesions. Spillover effects were reduced to less than 5% for hot spots > or = 3 x FWHM, at the expense of moderate noise amplification. CONCLUSION: While subtraction of object scatter is necessary for contrast enhancement and quantitation accuracy, restoration of detector scatter preserves sensitivity and improves quantitation accuracy by reducing spillover effects in high-resolution PET.
UNLABELLED: Although removal of object scatter has been shown to improve both contrast and quantitation accuracy, subtraction of detector scatter leads to marginal contrast enhancement and negligible resolution recovery at the expense of reduced sensitivity and increased statistical noise. Since detector scatter has correct information about radioactivity but slightly erroneous information about source location, we suggest that this component should be restored to preserve sensitivity and improve resolution. METHODS: A scatter correction model that consecutively removes object scatter and restores detector scatter is proposed. The scatter components are processed in the spatial domain using nonstationary scatter kernels. The detector scatter restoration kernel is obtained by piecewise inversion in the Fourier space. The model was tested using line source and hot spot phantom measurements. RESULTS: Object scatter subtraction increased contrast substantively with no effect on resolution. Detector scatter restoration recovered resolution almost completely with modest contrast enhancement in small lesions. Spillover effects were reduced to less than 5% for hot spots > or = 3 x FWHM, at the expense of moderate noise amplification. CONCLUSION: While subtraction of object scatter is necessary for contrast enhancement and quantitation accuracy, restoration of detector scatter preserves sensitivity and improves quantitation accuracy by reducing spillover effects in high-resolution PET.