TY - JOUR T1 - Spatial normalization of [<sup>18</sup>F]flutemetamol PET images utilizing an adaptive principal components template JF - Journal of Nuclear Medicine JO - J Nucl Med SP - 294 LP - 294 VL - 58 IS - supplement 1 AU - Johan Lilja AU - Antoine Leuzy AU - Konstantinos chiotis AU - Irina Savitcheva AU - Jens Sorensen AU - Agneta Nordberg Y1 - 2017/05/01 UR - http://jnm.snmjournals.org/content/58/supplement_1/294.abstract N2 - 294Objectives: Though currently validated for visual assessment only, there is evidence to suggest that quantitative approaches for use with 18F amyloid PET may prove advantageous. Quantification, typically in the form of SUVR, however, requires spatial normalization of PET images, a challenging task prone to bias due to different cortical uptake patterns. To overcome this, we here propose a method to spatially normalize [18F]flutemetamol images using a principal component based synthetic template.Methods: [18F]Flutemetamol PET and corresponding MR images from a phase II trial (n=70), including subjects ranging from amyloid negative to amyloid positive, were first spatially normalized to standard space using an MR driven registration method. [18F]Flutemetamol images were then intensity normalized to the pons. Principal component images in standard space were calculated from the intensity normalized [18F]flutemetamol images. A linear combination of the first two principal components was then used to model a synthetic template, spanning the range from amyloid negative to positive. This template was then incorporated in our registration method, where the optimal template was calculated as part of the registration process. In order to evaluate our method, we compared the quality of spatial normalization achieved with our method to that using an MR driven approach, including a validation cohort from an ongoing study investigating the clinical utility of amyloid PET in patients with an unclear diagnosis.Results: All scans were successfully spatially normalized using the proposed method, with no manual adjustments performed. Both visual and quantitative comparison between the PET and MR driven registration methods showed good agreement in cortical regions. [18F]flutemetamol quantification showed strong agreement between the SUVR values for the PET and MR driven methods (R2=0.996).Conclusion: The principal component template registration method allows for robust and accurate registration of [18F]flutemetamol images to a standardized template space, without the need for an MR image. Research Support: The [18F]flutemetamol phase II study was designed and sponsored by GE Healthcare. The majority of the [18F]flutemetamol data in the validation cohort was sponsored by Vinnova - the Swedish Governmental Agency for Innovation Systems, grant number 2013-05175. ER -