Objective: The limited resolution of PET imaging induces partial volume effects (PVE).The PVE makes the image quality impair.The aim of the study is to investigate the PVE and its influence factors for sphere lesion PET images with computer simulation experiments to provide the foundation for PVE correction.Methods: According to the theory of PVE formation, PET images of the sphere lesions with diameter 2mm-30mm were simulated with computer program in the ratio of background to lesion 0-0.7 under system resolutions 4mm-16mm.The maximum and the average pixel value and the full width at half maximum (the width at (maximum value-background)/2+background) of the lesion images were calculated.Results: The lesions sizes were indetermined on PET images.The full width at half maximum of the lesion image changed with real lesion size and system resolution and was independent of the background.The full width at half maximum of the lesion image was close to the real lesion size for the lesions with diameters more than the system resolution and close to the resolution size for the lesions with diameters less than the system resolution.The system resolution, real lesion size and intensity and background influenced the lesion images intensity.The maximum pixel value on the lesion image increased with increasing lesion diameter sharply for the lesions with diameters more than the double resolution and then increased slowly.The maximum pixel value reached 99% of the real lesion intensity for the lesions with diameters equal to the treble resolution.The average pixel value increased with increasing lesion diameter sharply for the lesions with diameters more than the treble resolution and then increased slowly.The difference of the average pixel values of lesion images among different backgrounds decreased with increasing lesion diameter.Conclusion: The real lesion size can be estimated with the full width of lesion image at half maximum and the real lesion intensity can be estimated with maximum or the average pixel value of lesion images, lesion size, system resolution and background.The background did not influence the full width of lesion image at half maximum and made the detecting power decline.