Effect of Point Spread Function Deconvolution in Reconstruction of Brain (18)F-FDG PET Images on the Diagnostic Thinking Efficacy in Alzheimer's Disease

Purpose: This study aims to determine the effect of applying Point Spread Function (PSF) deconvolution, which is known to improve contrast and spatial resolution in brain (18)F-FDG PET images, to the diagnostic thinking efficacy in Alzheimer's disease (AD). Methods: We compared Hoffman 3-D brain phantom images reconstructed with or without PSF. The effect of PSF deconvolution on AD diagnostic clinical performance was determined from digital brain (18)F-FDG PET images of AD (n = 38) and healthy (n = 35) subjects compared to controls (n = 36). Performances were assessed with SPM at the group level (p < 0.001 for the voxel) and at the individual level by visual interpretation of SPM T-maps (p < 0.005 for the voxel) by the consensual analysis of three experienced raters. Results: A mix of large hypometabolic (1,483cm(3), mean value of −867 ± 492 Bq/ml) and intense hypermetabolic (902 cm(3), mean value of 1,623 ± 1,242 Bq/ml) areas was observed in the PSF compared to the no PSF phantom images. Significant hypometabolic areas were observed in the AD group compared to the controls, for reconstructions with and without PSF (respectively 23.7 and 26.2 cm(3)), whereas no significant hypometabolic areas were observed when comparing the group of healthy subjects to the control group. At the individual level, no significant differences in diagnostic performances for discriminating AD were observed visually (sensitivity of 89 and 92% for reconstructions with and without PSF respectively, similar specificity of 74%). Conclusion: Diagnostic thinking efficacy performances for diagnosing AD are similar for (18)F-FDG PET images reconstructed with or without PSF.