PURPOSE: [(18)F]FEDAA1106 is expected to be used for evaluating the regional density of the peripheral benzodiazepine receptor (also called TSPO) in several neurodegenerative disorders. Regarding the quantification, direct binding potential (BP(ND)) has been reported to be preferable because of the variation of nondisplaceable distribution volume (V(ND)) among individuals. However, the precise calculation of BP(ND) is difficult in small regions or at voxel levels due to noise. Recently, a new graphical analysis (GA) was proposed to estimate V(ND) in a direct way. In this paper, we evaluated two types of GA for reliable quantification of BP(ND) in PET study with [(18)F]FEDAA1106 using computer simulations and human data. METHODS: In the simulations, time-activity curves were generated with various rate constants and noise levels, and the errors of BP(ND) estimated by GA were analyzed by comparing with true values calculated from rate constants given for the simulations. Thereafter, in a human study with [(18)F]FEDAA1106 for healthy volunteers, BP(ND) was estimated by two types of GA for region-of-interest (ROI) data. Parametric images of BP(ND) were generated by two types of GA with or without wavelet-denoising. RESULTS: Simulations showed that BP(ND) by GA was well correlated with true values, despite an underestimation. GA reduced unreasonable estimates compared with a conventional nonlinear least-square fitting (NLS), although larger variation of BP(ND) estimates was observed. In a ROI-based analysis of data obtained in a human study, BP(ND)s estimated by GA were well correlated with those generated by NLS, though they were underestimated. Parametric BP(ND) images by GA could be improved with wavelet-denoising. CONCLUSION: Graphical analysis could provide BP(ND) values with high stability and simple calculation in both ROI-based and voxel-based analyses of [(18)F]FEDAA1106 data.
PURPOSE: [(18)F]FEDAA1106 is expected to be used for evaluating the regional density of the peripheral benzodiazepine receptor (also called TSPO) in several neurodegenerative disorders. Regarding the quantification, direct binding potential (BP(ND)) has been reported to be preferable because of the variation of nondisplaceable distribution volume (V(ND)) among individuals. However, the precise calculation of BP(ND) is difficult in small regions or at voxel levels due to noise. Recently, a new graphical analysis (GA) was proposed to estimate V(ND) in a direct way. In this paper, we evaluated two types of GA for reliable quantification of BP(ND) in PET study with [(18)F]FEDAA1106 using computer simulations and human data. METHODS: In the simulations, time-activity curves were generated with various rate constants and noise levels, and the errors of BP(ND) estimated by GA were analyzed by comparing with true values calculated from rate constants given for the simulations. Thereafter, in a human study with [(18)F]FEDAA1106 for healthy volunteers, BP(ND) was estimated by two types of GA for region-of-interest (ROI) data. Parametric images of BP(ND) were generated by two types of GA with or without wavelet-denoising. RESULTS: Simulations showed that BP(ND) by GA was well correlated with true values, despite an underestimation. GA reduced unreasonable estimates compared with a conventional nonlinear least-square fitting (NLS), although larger variation of BP(ND) estimates was observed. In a ROI-based analysis of data obtained in a human study, BP(ND)s estimated by GA were well correlated with those generated by NLS, though they were underestimated. Parametric BP(ND) images by GA could be improved with wavelet-denoising. CONCLUSION: Graphical analysis could provide BP(ND) values with high stability and simple calculation in both ROI-based and voxel-based analyses of [(18)F]FEDAA1106 data.
Authors: Stergios Tsartsalis; Benjamin B Tournier; Christophe E Graf; Nathalie Ginovart; Vicente Ibáñez; Philippe Millet Journal: PLoS One Date: 2018-09-05 Impact factor: 3.240